Merge remote-tracking branch 'origin/master' into codex/rfc-subagent-background-tasks

Adopts #185 (dsh-timeout: clampTimeout/deadline/timeoutOf drive bash
run() timeout classification; runBash loses its own timer) and #108
(ask_user_question) across the task-runtime rework: bash-local keeps
the BashProcess handle shape with master's deadline mechanics, tool
catalogs/expectations carry both the task_* and ask-user tools, and
generated docs are regenerated on the union.

packages/README.md

@@ -11,16 +11,18 @@ Packages are grouped by modular role at `packages/<group>/<pkg>/`. The group dir
 | [`core/`](core/README.md) | Product API spine: session, system-prompt, tools, agent, and the concrete loop | Product — stable surface |
 | [`llm/`](llm/README.md) | LLM capability family: the abstract service + provider adapters | Product — stable surface |
 | [`bash/`](bash/README.md) | Bash capability family: the executor seam, a local impl, and the model-facing tool | Product — stable surface |
-| [`code-runtime/`](code-runtime/README.md) | Code-execution capability family: the abstract runtime seam for model-written programs | Product — stable surface |
+| [`code-runtime/`](code-runtime/README.md) | Code-execution capability family: the abstract runtime seam for model-written programs + a worker-thread backend | Product — stable surface |
 | [`fs/`](fs/README.md) | Filesystem capability family: the abstract seam, a local impl, and the model-facing file tools | Product — stable surface |
 | [`compact/`](compact/README.md) | Compaction capability family: the abstract seam + a basic backend (tool deferred) | Product — stable surface |
 | [`subagent/`](subagent/README.md) | Subagent capability family: the provider-registry seam and the model-facing delegation tool | Product — stable surface |
 | [`tasks/`](tasks/README.md) | Background task family: the `ctx.tasks` registry + the generic `task_*` control tools | Product — stable surface |
 | [`web/`](web/README.md) | Web capability family: the abstract seam, search/fetch provider impls, and the model-facing web tools | Product — stable surface |
-| [`todo/`](todo/README.md) | Todo/planning family: the model-facing `todo_write` tool (whole-list task tracking on the session log) | Product — stable surface |
+| [`timeout/`](timeout/README.md) | Tool-call timeout policy: the `tools/execute` deadline enforcer | Product — stable surface |
+| [`todo/`](todo/README.md) | Todo/planning family: the model-facing `todo_write` tool | Product — stable surface |
+| [`guard/`](guard/README.md) | Loop-hygiene guards: advisory repeat-call reminders | Product — stable surface |
 | [`hooks/`](hooks/README.md) | Hook bridges + the shared Claude Code / Codex wire-protocol library | Product — stable surface |
 | [`session-persistence/`](session-persistence/README.md) | Persistence capability family: the seam + JSONL/SQLite backends | Product — stable surface |
-| [`ui/`](ui/README.md) | Editor/client integration surfaces (the ACP bridge) + the app packages | Product — stable surface |
+| [`ui/`](ui/README.md) | Editor/client integration surfaces: ACP bridge, app packages, user-interaction seam, ask-user tool | Product — stable surface |
 | [`support/`](support/README.md) | Dev/test/example infrastructure (invariants, replay adapter, subagent mock) | Support — lower compatibility expectations |
 | [`util/`](util/README.md) | Low-level zero-dependency utilities shared across groups (the `Branded<B>` primitive) | Support — small, stable, harness-dep-free |
 
@@ -30,6 +32,6 @@ The split is the point: a package's group says whether it is part of the product
 
 The inter-package dependency graph is generated: [docs/module-graph.md](../docs/module-graph.md) (`pnpm run gen-module-graph`, freshness-gated in CI).
 
-The rule it must obey: **extension plugins depend on interfaces, never on the concrete loop.** `dsh-agent-loop` is swappable — UI/hook/tool plugins keep working against the `dsh-agent` vocabulary if the loop is replaced. The sanctioned exception is a **composition/bundle** package like `dsh-agent-core`, whose whole job is to assemble the concrete spine: it depends on `dsh-agent-loop` (and the other concrete spine plugins) on purpose. The rule constrains plugins that EXTEND the system, not the bundle that COMPOSES it — swapping the loop means shipping a different bundle, not rewiring every extension. A swappable capability splits into interface / implementation / consumer packages (the bash trio is the template — see [capability seams](../docs/rfc/implemented/architecture/2026-06-13-capability-seams.md)).
+The rule it must obey: **extension plugins depend on interfaces, never on the concrete loop.** `dsh-agent-loop` is swappable — UI/hook/tool plugins keep working against the `dsh-agent` vocabulary if the loop is replaced. The sanctioned exception is a **composition/bundle** package like `dsh-agent-core`, whose whole job is to assemble the concrete spine: it depends on `dsh-agent-loop` (and the other concrete spine plugins) on purpose. The rule constrains plugins that EXTEND the system, not the bundle that COMPOSES it. A swappable capability splits into interface / implementation / consumer packages (the bash trio is the template — see [capability seams](../docs/rfc/implemented/architecture/2026-06-13-capability-seams.md)).
 
 Each package has its own `README.md` with purpose, service API, events, extension points, and deliberate non-goals (TODOs).

packages/bash/bash-local/package.json

@@ -23,6 +23,7 @@
   "license": "BSD-3-Clause",
   "peerDependencies": {
     "@deepseek-ai/dsh-bash": "^0.0.1",
+    "@deepseek-ai/dsh-timeout": "^0.0.1",
     "cordis": "^4.0.0-rc.6"
   },
   "dependencies": {
@@ -30,6 +31,7 @@
   },
   "devDependencies": {
     "@deepseek-ai/dsh-bash": "workspace:^",
+    "@deepseek-ai/dsh-timeout": "workspace:^",
     "cordis": "^4.0.0-rc.6"
   }
 }

packages/bash/bash-local/src/index.ts

@@ -19,6 +19,7 @@ import { Context } from 'cordis'
 import z from 'schemastery'
 import { BashExecutor } from '@deepseek-ai/dsh-bash'
 import type { BashExecRequest, BashExecSpec, BashProcess, BashProcessRead, BashRunResult } from '@deepseek-ai/dsh-bash'
+import { clampTimeout, deadline, timeoutOf } from '@deepseek-ai/dsh-timeout'
 import { DEFAULT_GRACE_MS, runBash } from './run.ts'
 import type { RunInternals, RunningBash } from './run.ts'
 
@@ -107,8 +108,12 @@ export class LocalBashExecutor extends BashExecutor {
    * values and never re-default.
    */
   resolve(request: BashExecRequest): BashExecSpec {
-    if (request.timeoutMs !== undefined) assertPositiveFinite('request.timeoutMs', request.timeoutMs)
-    const timeoutMs = Math.min(request.timeoutMs ?? this.config.timeoutMs, this.config.maxTimeoutMs)
+    const timeoutMs = clampTimeout(
+      request.timeoutMs,
+      this.config.timeoutMs,
+      this.config.maxTimeoutMs,
+      'bash-local: request.timeoutMs',
+    )
     return {
       command: request.command,
       workdir: request.workdir ?? this.config.cwd ?? process.cwd(),
@@ -122,29 +127,39 @@ export class LocalBashExecutor extends BashExecutor {
   }
 
   async run(spec: BashExecSpec): Promise<BashRunResult> {
+    // One fused deadline drives both the timeout and upstream cancellation;
+    // runBash listens on d.signal and runs the SIGTERM→grace→SIGKILL kill.
+    // `using` clears the timer across the awaited process lifetime.
+    using d = deadline(spec.signal, spec.timeoutMs, 'BASH_TIMEOUT')
     const outcome = await runBash({
       command: spec.command,
       cwd: spec.workdir,
-      timeoutMs: spec.timeoutMs,
       maxOutputBytes: this.config.maxOutputBytes,
       graceMs: this.config.graceMs,
-      signal: spec.signal,
+      signal: d.signal,
       stdin: spec.stdin,
       env: spec.env,
     }, this.internals).done
-    return { ...outcome, timeoutMs: spec.timeoutMs }
+    // Classify the FIRST abort reason: a BASH_TIMEOUT TimeoutReason means our
+    // timeout cut the command short; any other abort — an upstream cancel, or a
+    // foreign (outer) deadline's timeout under nesting — is aborted. Scoping to
+    // our own code keeps a nested outer deadline from reading as our timeout.
+    // Mutually exclusive by construction — the fused signal reports one cause.
+    const timedOut = timeoutOf(d.signal, 'BASH_TIMEOUT') !== undefined
+    const aborted = d.signal.aborted && !timedOut
+    return { ...outcome, timedOut, aborted, timeoutMs: spec.timeoutMs }
   }
 
   start(spec: BashExecSpec): BashProcess {
     // No timeout for background processes (matches Claude Code, which
     // detaches the timeout when backgrounding); callers stop them via the
     // handle's kill() — or via spec.signal, which the seam contract honors
-    // for background runs too (runBash wires it to the group kill).
-    // spec.timeoutMs is ignored here by design.
+    // for background runs too (runBash wires it to the group kill). No
+    // deadline is created here, so spec.timeoutMs is ignored by design —
+    // background processes stay timeout-free (see the timeout-library RFC).
     const running = runBash({
       command: spec.command,
       cwd: spec.workdir,
-      timeoutMs: 0,
       maxOutputBytes: this.config.maxOutputBytes,
       graceMs: this.config.graceMs,
       signal: spec.signal,
@@ -160,8 +175,10 @@ export class LocalBashExecutor extends BashExecutor {
       exitCode: null,
       signal: null,
       done: running.done.then((outcome) => {
-        // Abort-killed processes report as killed, not completed.
-        if (proc.status === 'running') proc.status = outcome.aborted ? 'killed' : 'completed'
+        // Abort-killed processes report as killed, not completed. Background
+        // runs forward only the upstream signal (no timeout), so its aborted
+        // state is the authoritative "was this cancelled" signal.
+        if (proc.status === 'running') proc.status = spec.signal?.aborted === true ? 'killed' : 'completed'
         proc.exitCode = outcome.exitCode
         proc.signal = outcome.signal
         this.live.delete(proc)

packages/bash/bash-local/src/run.ts

@@ -6,6 +6,12 @@
  * Everything here is deliberately free of Cordis concepts so it can be unit
  * tested in isolation; `LocalBashExecutor` owns lifecycle and configuration.
  *
+ * runBash owns NO timing: it kills the process group when its `spec.signal`
+ * fires and does not distinguish a timeout from a cancel. The executor fuses
+ * timeout + upstream cancellation into that one signal via
+ * `@deepseek-ai/dsh-timeout`'s `deadline`, and classifies the outcome from the
+ * signal afterward — the timing/classification half is shared, the kill is not.
+ *
  * Design notes (surveyed against Claude Code, OpenCode, Codex, and pi — see
  * the package README): spawn-per-call with `detached: true` so the child
  * leads its own process group; kills target the group (`kill(-pid)`) so
@@ -71,13 +77,17 @@ export function childEnv(extra?: Record<string, string>): NodeJS.ProcessEnv {
 export interface SpawnSpec {
   command: string
   cwd: string
-  /** Kill the process group after this many milliseconds. 0 = no timeout. */
-  timeoutMs: number
   /** Per-stream in-memory cap; overflow spills to disk (tail kept in memory). */
   maxOutputBytes: number
   /** Grace period between the SIGTERM and the SIGKILL escalation on a kill. */
   graceMs: number
-  /** Abort signal — kills the process group when fired. */
+  /**
+   * Abort signal — kills the process group when it fires. The executor owns
+   * timing: `run()` passes a fused timeout/cancel deadline signal (see
+   * `@deepseek-ai/dsh-timeout`), `start()` passes the bare upstream signal.
+   * runBash only listens and kills; it does NOT classify why (the executor
+   * reads the signal's reason afterward).
+   */
   signal?: AbortSignal | undefined
   /**
    * Bytes to write to the child's stdin, then close it. Absent (or empty)
@@ -94,12 +104,15 @@ export interface SpawnSpec {
   env?: Record<string, string> | undefined
 }
 
-/** Raw outcome of one closed process (before result shaping). */
+/**
+ * Raw outcome of one closed process (before result shaping). Deliberately
+ * carries NO timeout/cancel classification: runBash kills on abort but does not
+ * decide why — the executor's `run()`/`start()` reads the deadline signal it
+ * owns to classify `timedOut`/`aborted` (see the package README).
+ */
 export interface SpawnOutcome {
   exitCode: number | null
   signal: NodeJS.Signals | null
-  timedOut: boolean
-  aborted: boolean
   stdout: CollectedOutput
   stderr: CollectedOutput
 }
@@ -343,9 +356,6 @@ export function runBash(spec: SpawnSpec, internals: RunInternals = {}): RunningB
   child.stdout.on('data', (chunk: Buffer) => { stdout.push(chunk) })
   child.stderr.on('data', (chunk: Buffer) => { stderr.push(chunk) })
 
-  let timedOut = false
-  let aborted = false
-  let killTimer: NodeJS.Timeout | undefined
   let graceTimer: NodeJS.Timeout | undefined
 
   // pid is undefined when the spawn itself fails (bad cwd, missing binary);
@@ -358,17 +368,12 @@ export function runBash(spec: SpawnSpec, internals: RunInternals = {}): RunningB
     graceTimer = setTimeout(() => { killGroup(pid, 'SIGKILL') }, spec.graceMs)
   }
 
-  if (spec.timeoutMs > 0) {
-    killTimer = setTimeout(() => {
-      timedOut = true
-      kill()
-    }, spec.timeoutMs)
-  }
-
-  const onAbort = (): void => {
-    aborted = true
-    kill()
-  }
+  // runBash owns no timer: the executor's `run()` fuses timeout+cancel into one
+  // deadline signal (`@deepseek-ai/dsh-timeout`) and passes it here; we only
+  // listen and run the SIGTERM→grace→SIGKILL kill. Whether the abort was a
+  // timeout or an upstream cancel is classified by the executor from that
+  // signal, not tracked here.
+  const onAbort = (): void => { kill() }
   spec.signal?.addEventListener('abort', onAbort, { once: true })
 
   // Write stdin and close it, but ONLY when the caller supplied bytes — with no
@@ -401,14 +406,11 @@ export function runBash(spec: SpawnSpec, internals: RunInternals = {}): RunningB
       resolve({
         exitCode,
         signal,
-        timedOut,
-        aborted,
         stdout: stdout.finalize(),
         stderr: stderr.finalize(),
       })
     })
     function cleanup(): void {
-      if (killTimer !== undefined) clearTimeout(killTimer)
       if (graceTimer !== undefined) clearTimeout(graceTimer)
       spec.signal?.removeEventListener('abort', onAbort)
     }

packages/bash/bash-local/tests/executor.spec.ts

@@ -77,6 +77,8 @@ describe('LocalBashExecutor.run', () => {
     const { bash } = await setup({ timeoutMs: 60_000 })
     const result = await bash.run(bash.resolve({ command: 'sleep 60', timeoutMs: 100 }))
     expect(result.timedOut).toBe(true)
+    // Mutually exclusive: a timeout classifies as timedOut, never also aborted.
+    expect(result.aborted).toBe(false)
     expect(result.timeoutMs).toBe(100)
   })
 
@@ -87,6 +89,20 @@ describe('LocalBashExecutor.run', () => {
     setTimeout(() => { controller.abort() }, 50)
     const result = await pending
     expect(result.aborted).toBe(true)
+    // Mutually exclusive: an upstream cancel classifies as aborted, never also timedOut.
+    expect(result.timedOut).toBe(false)
+  })
+
+  it('classifies a self-killed command as neither timed out nor aborted', async () => {
+    // The command kills itself (SIGTERM) with no timeout and no upstream abort:
+    // the deadline signal never fires, so both classifications are false — the
+    // fused-signal classification reports the cause that cut the command short,
+    // and here nothing the executor owns did.
+    const { bash } = await setup({ timeoutMs: 60_000 })
+    const result = await bash.run(bash.resolve({ command: 'kill -TERM $$' }))
+    expect(result.signal).toBe('SIGTERM')
+    expect(result.timedOut).toBe(false)
+    expect(result.aborted).toBe(false)
   })
 
   it('rejects on spawn failure (bad workdir)', async () => {

packages/bash/bash-local/tests/run.spec.ts

@@ -26,7 +26,6 @@ function spec(command: string, overrides: Partial<Parameters<typeof runBash>[0]>
   return {
     command,
     cwd: process.cwd(),
-    timeoutMs: 0,
     maxOutputBytes: 64_000,
     graceMs: 3_000,
     ...overrides,
@@ -56,13 +55,25 @@ async function waitForStdout(running: RunningBash, expected: string, timeoutMs =
   throw new Error(`stdout did not include ${JSON.stringify(expected)} after ${timeoutMs}ms`)
 }
 
+async function waitForPidFile(path: string, timeoutMs = 5_000): Promise<number> {
+  const deadline = Date.now() + timeoutMs
+  while (Date.now() < deadline) {
+    try {
+      const pid = Number(readFileSync(path, 'utf8').trim())
+      if (Number.isSafeInteger(pid) && pid > 0) return pid
+    } catch {
+      // The child shell has not written the pid file yet.
+    }
+    await new Promise(resolve => setTimeout(resolve, 20))
+  }
+  throw new Error(`pid file ${path} was not written after ${timeoutMs}ms`)
+}
+
 describe('runBash', () => {
   it('captures stdout on success', async () => {
     const result = await runBash(spec('echo hello')).done
     expect(result.exitCode).toBe(0)
     expect(result.signal).toBeNull()
-    expect(result.timedOut).toBe(false)
-    expect(result.aborted).toBe(false)
     expect(result.stdout.text).toBe('hello\n')
     expect(result.stdout.truncated).toBe(false)
     expect(result.stderr.text).toBe('')
@@ -97,17 +108,22 @@ describe('runBash', () => {
     expect(result.stdout.text.trim()).toMatch(/\/tmp$/)
   })
 
-  it('kills with SIGTERM on timeout', async () => {
+  it('kills the process group with SIGTERM when the signal fires', async () => {
+    // runBash owns no timer: it kills on abort. The executor drives the timeout
+    // by firing this signal via a deadline (see executor.spec.ts); here we
+    // assert the kill itself lands as SIGTERM.
+    const controller = new AbortController()
     const start = Date.now()
-    const result = await runBash(spec('sleep 60', { timeoutMs: 100 })).done
+    const running = runBash(spec('sleep 60', { signal: controller.signal }))
+    setTimeout(() => { controller.abort('deadline') }, 100)
+    const result = await running.done
     expect(Date.now() - start).toBeLessThan(5_000)
-    expect(result.timedOut).toBe(true)
     expect(result.signal).toBe('SIGTERM')
     expect(result.exitCode).toBeNull()
   })
 
   it('escalates to SIGKILL when SIGTERM is trapped', async () => {
-    const running = runBash(spec('trap \'\' TERM; echo ready; sleep 60', { graceMs: 200 }))
+    const running = runBash(spec('trap \'\' TERM; echo ready; while :; do sleep 60 & wait $!; done', { graceMs: 200 }))
     await waitForStdout(running, 'ready\n')
     running.kill()
     const result = await running.done
@@ -119,8 +135,7 @@ describe('runBash', () => {
     // group must take the sleep down with bash.
     const pidFile = join(spillDir, `grandchild-${Date.now()}.pid`)
     const running = runBash(spec(`sleep 60 & echo $! > ${pidFile}; wait`))
-    await new Promise(resolve => setTimeout(resolve, 300))
-    const grandchild = Number(readFileSync(pidFile, 'utf8').trim())
+    const grandchild = await waitForPidFile(pidFile)
     expect(grandchild).toBeGreaterThan(0)
 
     running.kill()
@@ -134,7 +149,6 @@ describe('runBash', () => {
     const running = runBash(spec('sleep 60', { signal: controller.signal }))
     setTimeout(() => { controller.abort('user cancelled') }, 50)
     const result = await running.done
-    expect(result.aborted).toBe(true)
     expect(result.signal).toBe('SIGTERM')
   })
 
@@ -211,7 +225,6 @@ describe('stdin and extra env (set by in-process plugins)', () => {
     const big = 'x'.repeat(1024 * 1024)
     const result = await runBash(spec('exit 7', { stdin: big })).done
     expect(result.exitCode).toBe(7)
-    expect(result.aborted).toBe(false)
   })
 })
 
@@ -339,11 +352,11 @@ describe('abort edge cases', () => {
       .toThrow(/aborted before spawn: aborted/)
   })
 
-  it('reports an externally self-killed command without the timeout marker', async () => {
+  it('reports the terminating signal of an externally self-killed command', async () => {
+    // runBash reports the raw signal; whether it counts as timeout/cancel is the
+    // executor's classification (a self-kill is neither) — see executor.spec.ts.
     const result = await runBash(spec('kill -TERM $$')).done
     expect(result.signal).toBe('SIGTERM')
-    expect(result.timedOut).toBe(false)
-    expect(result.aborted).toBe(false)
   })
 })
 
@@ -396,10 +409,9 @@ describe('review fixes: env scrubbing and spill hardening', () => {
 
   it('honors AbortSignal on background-style runs (no timeout)', async () => {
     const controller = new AbortController()
-    const running = runBash(spec('sleep 60', { timeoutMs: 0, signal: controller.signal }))
+    const running = runBash(spec('sleep 60', { signal: controller.signal }))
     setTimeout(() => { controller.abort() }, 50)
     const result = await running.done
-    expect(result.aborted).toBe(true)
     expect(result.signal).toBe('SIGTERM')
   })
 })

packages/bash/bash-local/tsconfig.json

@@ -20,6 +20,9 @@
     {
       "path": "../../util/brand"
     },
+    {
+      "path": "../../util/timeout"
+    },
     {
       "path": "../../bash/bash"
     }

packages/code-runtime/README.md

@@ -1,9 +1,10 @@
 # code-runtime/ — code-execution capability family
 
-The code-execution capability seam (see [capability seams](../../docs/rfc/implemented/architecture/2026-06-13-capability-seams.md)): an abstract runtime interface for executing one model-written program against host-provided async bindings, capturing what it printed and returned. The consumer is the tool registry's Code Mode, and the first implementation (a Node worker-thread backend) is specified alongside it in the [Code Mode RFC](../../docs/rfc/proposed/feature/2026-06-15-code-mode.md). **Product** packages.
+The code-execution capability seam (see [capability seams](../../docs/rfc/implemented/architecture/2026-06-13-capability-seams.md)): an abstract runtime interface for executing one model-written program against host-provided async bindings, capturing what it printed and returned. The consumer is the tool registry's Code Mode, specified alongside the seam in the [Code Mode RFC](../../docs/rfc/proposed/feature/2026-06-15-code-mode.md). **Product** packages.
 
 | Package | Role | ctx key |
 |---|---|---|
 | `code-runtime/` | Abstract code-execution seam (interface + vocabulary) | `ctx.codeRuntime` |
+| [`code-runtime-worker/`](code-runtime-worker/README.md) | Worker-thread backend: fresh worker per run, TypeScript via host-side type-strip (annotations advisory, never type-checked), port-bridged bindings, budget/heap containment | registers `ctx.codeRuntime` |
 
-The interface lives at `code-runtime/code-runtime/`. Backends differ by execution substrate (worker thread, process, container) and by source language — both readonly descriptors on the service — and register `ctx.codeRuntime` without touching the interface or its consumer; that split is what makes a hardened backend a drop-in later.
+The interface lives at `code-runtime/code-runtime/`; the shipped backend at `code-runtime/code-runtime-worker/`. Backends differ by execution substrate (worker thread, process, container) and by source language — both readonly descriptors on the service — and register `ctx.codeRuntime` without touching the interface or its consumer; that split is what makes a hardened backend a drop-in later.

packages/code-runtime/code-runtime-worker/README.md

@@ -0,0 +1,32 @@
+# @deepseek-ai/dsh-code-runtime-worker
+
+Worker-thread implementation of the [`@deepseek-ai/dsh-code-runtime`](../code-runtime/README.md) seam: `WorkerCodeRuntime` runs each program in ONE fresh Node `worker_threads.Worker` — TypeScript in, type-stripped host-side, bindings bridged over the message port, `{ value, logs, error? }` out. **Containment, not a security boundary**: trust posture is bash-equivalent by design (the [Code Mode RFC](../../../docs/rfc/proposed/feature/2026-06-15-code-mode.md) § Trust posture), with containment bash does not have — separate isolate, empty environment, heap cap, hard termination.
+
+## Config
+
+```yaml
+- id: code-runtime
+  name: '@deepseek-ai/dsh-code-runtime-worker'
+  config:
+    computeMs: 60000              # busy-time budget (measured event-loop active time)
+    maxWallMs: 600000             # wall-clock ceiling; never pauses for anything
+    maxLogBytes: 65536            # shared byte budget for captured log text
+    maxValueBytes: 32768          # rendered-completion-value cap
+    maxOldGenerationSizeMb: 512   # worker heap cap (resourceLimits)
+```
+
+Every field is validated (positive numbers) and defaulted; there are no other tunables.
+
+## Design
+
+- **One fresh worker per run, no pooling** — a program's world dies with its worker: no cross-run state to log, state bleed unrepresentable, runs reconstructable from the session log alone.
+- **Type-strip host-side, in execution context** — the program is wrapped in an async-function shell, stripped with `node:module`'s `stripTypeScriptTypes` (erasable syntax only — `enum`/namespaces are rejected as a program `exception` and no worker spawns), and sliced back out byte-positioned; it then executes as the body of an `AsyncFunction`, so top-level `await`/`return` work.
+- **The port assumes a hostile peer** — model code can reach `parentPort` and forge traffic, so every inbound message is shape-validated and REBUILT before anything reads it (`null`, primitives, junk types, and malformed payloads drop without a throw; forged extra fields never ride along), the host answers each call id at most once, resolves binding names as OWN properties only (a forged `constructor` cannot walk a prototype chain), drops post-settlement replies, and converts a non-cloneable binding resolution into an error reply. Forged `log`/`done` messages cannot bypass the caps: one host-side ledger bounds everything that lands in `logs`, and the completion value is re-capped host-side. Worker-side namespaces are null-prototype with `defineProperty`, so `__proto__`-shaped binding names are ordinary keys.
+- **Two independent budgets, because the peer is hostile** — `computeMs` meters the worker's MEASURED busy time (`worker.performance.eventLoopUtilization()` polling): a hot loop cannot hide behind a pending decoy dispatch, and a program awaiting a slow tool accrues nothing. `maxWallMs` backstops what busy time cannot see (awaiting a promise nobody resolves). Both funnel into `worker.terminate()`, which ends hot synchronous loops too; heap overflow surfaces as the worker's OOM exit (`kind: 'worker-exit'`).
+- **Logs stream eagerly** — console/stdout/stderr entries cross the port as they happen, so a timed-out or killed program still shows what it printed. ONE shared `maxLogBytes` ledger bounds everything: streamed entries, forged port traffic, and pipe bytes that bypass the patched streams (appended after), with the overflow marked in-band once.
+- **Empty environment** — the worker gets `env: {}` and `execArgv: []`: no ambient credentials (stronger than the scrubbed-env rule for spawned commands) and no inherited loader flags.
+- **Dispose to quiescence** — teardown fails in-flight runs as `abort` and AWAITS each worker's exit before resolving.
+
+## The worker entry, unbuilt and built
+
+`worker.ts` is deliberately erasable-only TypeScript with type-only cross-package imports: unbuilt (vitest/tsx), the host spawns `src/worker.ts` directly and Node's native type stripping loads it; built, the entry ships as the sibling bundle `lib/worker.js` (its own tsdown entry). The built path is pinned by `tests/built-lib.e2e.ts`, the real-load-path guard from [docs/testing.md](../../../docs/testing.md).

packages/code-runtime/code-runtime-worker/package.json

@@ -0,0 +1,36 @@
+{
+  "name": "@deepseek-ai/dsh-code-runtime-worker",
+  "description": "Worker-thread implementation of the DeepSeek Harness code-execution seam",
+  "version": "0.0.1",
+  "private": true,
+  "type": "module",
+  "main": "lib/index.js",
+  "types": "lib/types/index.d.ts",
+  "exports": {
+    ".": {
+      "types": "./lib/types/index.d.ts",
+      "default": "./lib/index.js"
+    },
+    "./src/*": "./src/*",
+    "./package.json": "./package.json"
+  },
+  "files": [
+    "lib/index.js",
+    "lib/worker.js",
+    "lib/types/**/*.d.ts",
+    "lib/types/**/*.d.ts.map",
+    "src"
+  ],
+  "license": "BSD-3-Clause",
+  "peerDependencies": {
+    "@deepseek-ai/dsh-code-runtime": "^0.0.1",
+    "cordis": "^4.0.0-rc.6"
+  },
+  "dependencies": {
+    "schemastery": "^3.18.0"
+  },
+  "devDependencies": {
+    "@deepseek-ai/dsh-code-runtime": "workspace:^",
+    "cordis": "^4.0.0-rc.6"
+  }
+}

packages/code-runtime/code-runtime-worker/src/bootstrap.ts

@@ -0,0 +1,301 @@
+/**
+ * Worker-side execution logic, written as plain functions over an injected
+ * port so the unit suite can run every line IN-PROCESS against a fake port
+ * (a real worker thread is a separate V8 isolate the coverage provider
+ * cannot observe). The real worker entry (`worker.ts`) is a thin
+ * self-executing glue file over {@link runWorkerMain}, excluded from
+ * coverage the same way `bin.ts` entrypoints are, and exercised end-to-end
+ * by the integration tests that spawn real workers.
+ *
+ * @module @deepseek-ai/dsh-code-runtime-worker/src/bootstrap
+ */
+
+import { inspect } from 'node:util'
+import { serialize } from 'node:v8'
+import type { CodeLogEntry } from '@deepseek-ai/dsh-code-runtime'
+import { logTruncationMarker } from './protocol.ts'
+import type { DoneMessage, ReplyMessage, WorkerBootData, WorkerToHost } from './protocol.ts'
+
+/** The port surface the bootstrap needs — satisfied by `parentPort` and by the tests' fake. */
+export interface BootstrapPort {
+  postMessage(message: WorkerToHost): void
+  on(event: 'message', listener: (message: ReplyMessage) => void): void
+}
+
+/**
+ * A writable stream's `write` slot, as the bootstrap patches it (see
+ * {@link captureStreamWrites}). Method-typed so the real
+ * `process.stdout`/`process.stderr` (narrower chunk parameters) remain
+ * assignable.
+ */
+export interface PatchableStream {
+  write(chunk: unknown, ...rest: unknown[]): boolean
+}
+
+/**
+ * Ordered log capture under one shared byte budget, delivered to a sink as
+ * each entry lands (the real sink streams entries over the port eagerly, so
+ * captured output survives a mid-run termination). Once the budget is
+ * exhausted it emits exactly one in-band marker entry (on the `stderr`
+ * diagnostics channel) and silently drops everything after — the cap is a
+ * blast-radius bound, so "how much was lost" intentionally stays unmeasured.
+ */
+export class LogBuffer {
+  private remaining: number
+  private truncated = false
+  // Explicit fields, not constructor parameter properties: this module loads
+  // under Node's native strip-only mode, which rejects non-erasable syntax —
+  // and parameter properties are non-erasable.
+  private readonly maxBytes: number
+  private readonly sink: (entry: CodeLogEntry) => void
+
+  constructor(maxBytes: number, sink: (entry: CodeLogEntry) => void) {
+    this.maxBytes = maxBytes
+    this.sink = sink
+    this.remaining = maxBytes
+  }
+
+  /**
+   * Emit one entry to the sink, charging its text against the budget (drops + marks once exhausted).
+   * @param entry - the log entry to deliver.
+   */
+  push(entry: CodeLogEntry): void {
+    if (this.truncated) return
+    const cost = Buffer.byteLength(entry.text, 'utf8')
+    if (cost > this.remaining) {
+      this.truncated = true
+      this.sink({ source: 'stderr', text: logTruncationMarker(this.maxBytes) })
+      return
+    }
+    this.remaining -= cost
+    this.sink(entry)
+  }
+}
+
+/** The five console methods the shim captures, in the seam's level vocabulary. */
+const CONSOLE_LEVELS = ['log', 'info', 'warn', 'error', 'debug'] as const
+
+/**
+ * A `console` replacement whose five leveled methods render their arguments
+ * `util.inspect`-style (matching real console formatting closely enough for
+ * a model to recognize its own output) into the buffer. Only these five
+ * exist — the program gets a deliberately small console, not Node's full
+ * surface.
+ * @param logs - the buffer every rendered line is pushed into.
+ * @returns the five-method console object handed to the program.
+ */
+export function makeConsoleShim(logs: LogBuffer): Record<(typeof CONSOLE_LEVELS)[number], (...args: unknown[]) => void> {
+  const render = (args: unknown[]): string =>
+    args.map(arg => typeof arg === 'string' ? arg : inspect(arg, INSPECT_OPTIONS)).join(' ')
+  const shim = Object.create(null) as Record<(typeof CONSOLE_LEVELS)[number], (...args: unknown[]) => void>
+  for (const level of CONSOLE_LEVELS) {
+    shim[level] = (...args: unknown[]) => { logs.push({ source: 'console', level, text: render(args) }) }
+  }
+  return shim
+}
+
+/**
+ * Redirect a stream's `write` into the log buffer (the program-visible
+ * `process.stdout`/`process.stderr` in the real worker), so raw writes land
+ * in emission order alongside console output instead of racing down a pipe.
+ * The shim keeps Node's `write(chunk[, encoding][, callback])` contract: the
+ * callback fires asynchronously once the chunk is admitted (a program
+ * awaiting flush completion must complete, not sit until the wall timeout),
+ * even for writes the exhausted budget drops.
+ * @param logs - the buffer captured writes are pushed into.
+ * @param stream - the stream whose `write` slot is patched.
+ * @param source - the log source the captured writes are attributed to.
+ * @returns the restore function (the in-process tests un-patch; the real
+ *   worker never needs to).
+ */
+export function captureStreamWrites(logs: LogBuffer, stream: PatchableStream, source: 'stdout' | 'stderr'): () => void {
+  // The slot's VALUE is stored for restore and reassigned — never invoked
+  // detached, so the unbound-method concern does not apply.
+  // eslint-disable-next-line @typescript-eslint/unbound-method
+  const original = stream.write
+  stream.write = (chunk: unknown, ...rest: unknown[]): boolean => {
+    logs.push({ source, text: typeof chunk === 'string' ? chunk : String(chunk) })
+    // Node's optional-encoding shape: the callback is whichever of the next
+    // two positions holds a function (a non-function there is the encoding).
+    const callback = [rest[0], rest[1]].find(
+      (arg): arg is (error?: Error | null) => void => typeof arg === 'function',
+    )
+    if (callback) queueMicrotask(() => { callback(null) })
+    return true
+  }
+  return () => { stream.write = original }
+}
+
+/** Bounded inspect options: deep enough to be useful, bounded so a pathological value cannot explode the rendering. */
+const INSPECT_OPTIONS = { depth: 4, maxArrayLength: 100, maxStringLength: 10_000 } as const
+
+/**
+ * The longest prefix of `text` whose UTF-8 encoding fits `maxBytes`, cut at
+ * a code-point boundary (never mid-surrogate-pair). The byte caps are BYTE
+ * caps — `String.prototype.slice` counts UTF-16 code units, up to 3× smaller
+ * than what a multibyte string actually costs across the boundary.
+ * @param text - the string to bound.
+ * @param maxBytes - the UTF-8 byte budget the prefix must fit.
+ * @returns the prefix (all of `text` when it already fits).
+ */
+export function truncateUtf8Bytes(text: string, maxBytes: number): string {
+  if (Buffer.byteLength(text, 'utf8') <= maxBytes) return text
+  let bytes = 0
+  let end = 0
+  for (const char of text) {
+    const cost = Buffer.byteLength(char, 'utf8')
+    if (bytes + cost > maxBytes) break
+    bytes += cost
+    end += char.length
+  }
+  return text.slice(0, end)
+}
+
+/**
+ * Prepare the program's completion value for the done message: a value whose
+ * MEASURED cross-boundary size fits `maxValueBytes` crosses raw — exact
+ * bytes for a string, the structured-clone wire size (`v8.serialize`) for
+ * everything else, so a huge container whose BOUNDED inspect rendering
+ * happens to be small cannot smuggle itself past the cap. Anything else
+ * (non-cloneable, or oversized) is REPLACED by its bounded `util.inspect`
+ * rendering, byte-truncated ({@link truncateUtf8Bytes}) with an in-band
+ * marker — the seam contract's "a non-transferable value is replaced by a
+ * string rendering", extended to oversized ones so a huge return cannot
+ * flood the host.
+ * @param value - the program's completion value.
+ * @param maxValueBytes - the byte cap for the value.
+ * @returns the done-message fragment: `{}` for `undefined`, else `{ value }`.
+ */
+export function prepareValue(value: unknown, maxValueBytes: number): { value?: unknown } {
+  if (value === undefined) return {}
+  if (typeof value === 'string') {
+    if (Buffer.byteLength(value, 'utf8') <= maxValueBytes) return { value }
+  } else {
+    let size: number | undefined
+    try {
+      size = serialize(value).byteLength
+    } catch {
+      // Only the verdict matters: the value has parts the structured-clone
+      // algorithm rejects (functions, classes, …) and must cross as its
+      // rendering instead.
+      size = undefined
+    }
+    if (size !== undefined && size <= maxValueBytes) return { value }
+  }
+  const rendered = typeof value === 'string' ? value : inspect(value, INSPECT_OPTIONS)
+  const capped = Buffer.byteLength(rendered, 'utf8') > maxValueBytes
+    ? `${truncateUtf8Bytes(rendered, maxValueBytes)}… [truncated]`
+    : rendered
+  return { value: capped }
+}
+
+/** One awaited binding call's settlement handles, keyed by call id in the pending map. */
+export interface PendingCall {
+  resolve(value: unknown): void
+  reject(error: Error): void
+}
+
+/**
+ * Route host replies into the pending-call map: each reply settles its call
+ * at most once, and a reply for an unknown id (stray, or a duplicate answer
+ * to an id already settled) is ignored. Shared wiring between
+ * {@link runWorkerMain} and the tests that exercise {@link makeNamespaces}
+ * standalone.
+ * @param port - the port whose `message` events carry the replies.
+ * @param pending - the id-keyed map of unsettled binding calls.
+ */
+export function wireReplies(port: BootstrapPort, pending: Map<number, PendingCall>): void {
+  port.on('message', (message: ReplyMessage) => {
+    const entry = pending.get(message.id)
+    if (!entry) return
+    pending.delete(message.id)
+    if (message.ok) entry.resolve(message.value)
+    else entry.reject(new Error(message.message))
+  })
+}
+
+/**
+ * Build the binding namespace objects the program sees: one null-prototype
+ * global per namespace, each declared name an own enumerable async function
+ * that bridges over the port (`__proto__`/`constructor`/`toString` are
+ * ordinary keys, never prototype collisions). A non-cloneable argument
+ * rejects that one call with a descriptive error; the host's reply (`ok`
+ * false) rejects it likewise, so a failed tool call surfaces in the program
+ * as an ordinary promise rejection.
+ * @param data - the boot payload's namespace declarations (globals + names).
+ * @param port - the port binding calls are posted to.
+ * @param pending - the id-keyed map each posted call parks its handles in.
+ * @param nextId - the shared mutable id counter (worker-issued correlation ids).
+ * @returns one namespace object per declaration, in declaration order.
+ */
+export function makeNamespaces(
+  data: Pick<WorkerBootData, 'namespaces'>,
+  port: BootstrapPort,
+  pending: Map<number, PendingCall>,
+  nextId: { value: number },
+): Record<string, unknown>[] {
+  return data.namespaces.map(({ global, names }) => {
+    const namespace = Object.create(null) as Record<string, unknown>
+    for (const name of names) {
+      Object.defineProperty(namespace, name, {
+        enumerable: true,
+        value: (args: unknown): Promise<unknown> => new Promise((resolve, reject) => {
+          const id = nextId.value++
+          pending.set(id, { resolve, reject })
+          try {
+            port.postMessage({ type: 'call', id, global, name, args })
+          } catch (error: unknown) {
+            pending.delete(id)
+            reject(new Error(`binding arguments must be structured-cloneable: ${error instanceof Error ? error.message : String(error)}`))
+          }
+        }),
+      })
+    }
+    return namespace
+  })
+}
+
+/**
+ * Run one program to settlement and post the {@link DoneMessage}: wires the
+ * reply handler, materializes the namespaces and console shim, compiles the
+ * type-stripped body as an async function (top-level `await`/`return`
+ * work), and reports a thrown program error as the done message's `error`
+ * field. Exactly one done message is ever posted.
+ * @param port - the message port to the host (the real `parentPort`, or the tests' fake).
+ * @param data - the boot payload the host sent.
+ * @param streams - the stream objects whose `write` is captured (the real
+ *   `process.stdout`/`process.stderr` in the worker; fakes in tests).
+ * @returns resolves after the done message is posted (the tests await it;
+ *   the real entry lets the worker exit naturally).
+ */
+export async function runWorkerMain(
+  port: BootstrapPort,
+  data: WorkerBootData,
+  streams: { stdout: PatchableStream; stderr: PatchableStream },
+): Promise<void> {
+  const logs = new LogBuffer(data.maxLogBytes, (entry) => { port.postMessage({ type: 'log', entry }) })
+  captureStreamWrites(logs, streams.stdout, 'stdout')
+  captureStreamWrites(logs, streams.stderr, 'stderr')
+
+  const pending = new Map<number, PendingCall>()
+  wireReplies(port, pending)
+
+  const nextId = { value: 1 }
+  const namespaces = makeNamespaces(data, port, pending, nextId)
+  const consoleShim = makeConsoleShim(logs)
+
+  let done: DoneMessage
+  try {
+    // The async function constructor, reached through an instance because
+    // `AsyncFunction` is not a global. The program body is strict-mode.
+    /* v8 ignore next -- the arrow exists only to reach the AsyncFunction constructor; it is never invoked. */
+    const AsyncFunction = (async () => {}).constructor as new (...args: string[]) => (...fnArgs: unknown[]) => Promise<unknown>
+    const fn = new AsyncFunction(...data.namespaces.map(namespace => namespace.global), 'console', `'use strict';\n${data.code}`)
+    const value = await fn(...namespaces, consoleShim)
+    done = { type: 'done', ...prepareValue(value, data.maxValueBytes) }
+  } catch (error: unknown) {
+    const message = error instanceof Error ? error.stack ?? error.message : String(error)
+    done = { type: 'done', error: { message } }
+  }
+  port.postMessage(done)
+}

packages/code-runtime/code-runtime-worker/src/index.ts

@@ -0,0 +1,441 @@
+/**
+ * Worker-thread implementation of the code-execution seam: one fresh Node
+ * worker per run, executing the model's TypeScript after a host-side
+ * type-strip, with bindings bridged over the message port. Containment, not
+ * a security boundary (bash-equivalent trust — see the Code Mode RFC's
+ * trust-posture section): the worker gets an EMPTY environment, a heap cap,
+ * and two independent budgets — `computeMs` metered on the worker's
+ * measured event-loop busy time (a hot loop cannot hide behind a pending
+ * binding call) and a never-pausing `maxWallMs` ceiling — all funneling
+ * into `worker.terminate()`, which ends hot synchronous loops too.
+ *
+ * @module @deepseek-ai/dsh-code-runtime-worker
+ */
+
+import { Worker } from 'node:worker_threads'
+import { stripTypeScriptTypes } from 'node:module'
+import { Context } from 'cordis'
+import z from 'schemastery'
+import { CodeRuntime } from '@deepseek-ai/dsh-code-runtime'
+import type { CodeBindingFunction, CodeLogEntry, CodeRunFailure, CodeRunRequest, CodeRunResult } from '@deepseek-ai/dsh-code-runtime'
+import { prepareValue, truncateUtf8Bytes } from './bootstrap.ts'
+import { logTruncationMarker } from './protocol.ts'
+import type { ReplyMessage, WorkerBootData, WorkerToHost } from './protocol.ts'
+
+export type { BootstrapPort, PatchableStream } from './bootstrap.ts'
+export type { CallMessage, DoneMessage, ReplyMessage, WorkerBootData, WorkerToHost } from './protocol.ts'
+
+/** Plugin config: every execution cap, changeable from `cordis.yml` (no hardcoded tunables). */
+export interface Config {
+  /**
+   * Busy-time budget in milliseconds: the run fails with kind `'timeout'`
+   * once the worker's MEASURED event-loop active time
+   * (`worker.performance.eventLoopUtilization()`) exceeds this. Metering
+   * measured busy time — not wall time, not host-side pending-call
+   * bookkeeping — is what makes the budget both fair (a program awaiting a
+   * slow tool accrues nothing) and ungameable (a hot loop accrues whether
+   * or not a decoy dispatch is in flight).
+   */
+  computeMs?: number
+  /**
+   * Wall-clock ceiling in milliseconds; never pauses for anything. The
+   * backstop for what busy-time cannot see (a program awaiting a promise
+   * nobody will resolve).
+   */
+  maxWallMs?: number
+  /** Shared byte budget for captured log text (console + raw stream writes), truncation marked in-band. */
+  maxLogBytes?: number
+  /**
+   * Byte cap for the completion value, measured by its real cross-boundary
+   * size (string bytes, or structured-clone wire size); an oversized or
+   * non-cloneable value crosses as a capped string rendering.
+   */
+  maxValueBytes?: number
+  /** The worker's max old-generation heap in MiB (`resourceLimits`); overflow kills the worker, surfacing as kind `'worker-exit'`. */
+  maxOldGenerationSizeMb?: number
+}
+
+/** {@link Config} after schemastery fills the defaults (every field present). */
+type ResolvedConfig = Required<Config>
+
+/**
+ * How often the host samples the worker's event-loop utilization for the
+ * `computeMs` budget. An internal cadence, not config: the only effect of
+ * the interval is budget-expiry granularity (a run can overshoot by up to
+ * one interval), and nothing a deployment could tune here improves that
+ * without burning host CPU.
+ */
+const ELU_POLL_INTERVAL_MS = 25
+
+/** ECMAScript reserved words that cannot be async-function parameter names — rejected as binding globals. */
+const RESERVED_WORDS = new Set([
+  'await', 'break', 'case', 'catch', 'class', 'const', 'continue', 'debugger', 'default', 'delete', 'do',
+  'else', 'enum', 'export', 'extends', 'false', 'finally', 'for', 'function', 'if', 'import', 'in',
+  'instanceof', 'new', 'null', 'return', 'super', 'switch', 'this', 'throw', 'true', 'try', 'typeof',
+  'var', 'void', 'while', 'with', 'yield', 'let', 'static', 'implements', 'interface', 'package',
+  'private', 'protected', 'public', 'arguments', 'eval',
+])
+
+/** Valid async-function parameter name (the binding global becomes one). */
+const IDENTIFIER = /^[A-Za-z_$][A-Za-z0-9_$]*$/
+
+/**
+ * The shell a program is wrapped in for the type-strip, matching the
+ * grammatical context it will execute in (an async function body, where
+ * top-level `return` and `await` are legal — a bare module parse would
+ * reject the `return`). Strip mode is position-preserving (removed syntax
+ * becomes whitespace, nothing shifts), so the wrapper survives the strip
+ * byte-identical and the body slices back out with the model's own
+ * line/column positions intact.
+ */
+const STRIP_WRAP = { prefix: 'async function __dsh_program__() {\n', suffix: '\n}' } as const
+
+/** One in-flight run's host-side state, tracked for disposal. */
+interface LiveRun {
+  worker: Worker
+  settle(failure: CodeRunFailure): void
+  finished: Promise<void>
+}
+
+/**
+ * The worker entry module. Source runs unbuilt (`src/worker.ts`, loadable
+ * directly on this repo's Node range via native type stripping — the file
+ * is erasable-only with type-only relative imports); the built package
+ * ships it as a sibling bundle (`lib/worker.js`, its own tsdown entry).
+ * The URL *pathname*'s extension says which world this module is in —
+ * pathname, because dev-time module runners (vitest) may suffix
+ * `import.meta.url` with a query string; relative resolution drops it.
+ */
+/* v8 ignore next -- the './worker.js' arm is the built-lib world, unreachable unbuilt by construction; the built-lib e2e pins it. */
+const WORKER_URL = new URL(new URL(import.meta.url).pathname.endsWith('.ts') ? './worker.ts' : './worker.js', import.meta.url)
+
+/** Render an unknown thrown value as a message, `Error` or not. */
+function messageOf(error: unknown): string {
+  return error instanceof Error ? error.message : String(error)
+}
+
+/** The log sources / console levels the seam vocabulary admits, as runtime sets for inbound-message validation. */
+const LOG_SOURCES = new Set<string>(['console', 'stdout', 'stderr'])
+const LOG_LEVELS = new Set<string>(['log', 'info', 'warn', 'error', 'debug'])
+
+/**
+ * Runtime shape gate for inbound port traffic. The peer runs MODEL CODE and
+ * can post anything — `null`, primitives, objects with poisoned fields — so
+ * the compile-time `WorkerToHost` type means nothing here: everything is
+ * re-validated and REBUILT field by field (a forged extra field never rides
+ * along; a non-number call id can never be echoed into a reply). Junk returns
+ * `undefined` and is dropped — a throw in the host's `message` listener would
+ * crash the host process.
+ */
+function parseWorkerMessage(raw: unknown): WorkerToHost | undefined {
+  if (typeof raw !== 'object' || raw === null) return undefined
+  const m = raw as Record<string, unknown>
+  switch (m.type) {
+    case 'call': {
+      if (typeof m.id !== 'number' || typeof m.global !== 'string' || typeof m.name !== 'string') return undefined
+      return { type: 'call', id: m.id, global: m.global, name: m.name, args: m.args }
+    }
+    case 'log': {
+      const entry = m.entry
+      if (typeof entry !== 'object' || entry === null) return undefined
+      const e = entry as Record<string, unknown>
+      if (typeof e.text !== 'string') return undefined
+      if (typeof e.source !== 'string' || !LOG_SOURCES.has(e.source)) return undefined
+      if (e.level !== undefined && (typeof e.level !== 'string' || !LOG_LEVELS.has(e.level))) return undefined
+      return {
+        type: 'log',
+        entry: {
+          source: e.source as CodeLogEntry['source'],
+          ...e.level !== undefined ? { level: e.level as Exclude<CodeLogEntry['level'], undefined> } : {},
+          text: e.text,
+        },
+      }
+    }
+    case 'done': {
+      if (m.error === undefined) return { type: 'done', ...m.value !== undefined ? { value: m.value } : {} }
+      const error = m.error
+      if (typeof error !== 'object' || error === null) return undefined
+      const message = (error as Record<string, unknown>).message
+      if (typeof message !== 'string') return undefined
+      return { type: 'done', ...m.value !== undefined ? { value: m.value } : {}, error: { message } }
+    }
+    default: return undefined
+  }
+}
+
+/**
+ * Headroom the host's value re-cap grants over `maxValueBytes`: exactly the
+ * truncation suffix {@link prepareValue} appends, so a value the WORKER
+ * already capped (byte-exact prefix + this marker) passes through unchanged
+ * instead of being marked twice.
+ */
+const VALUE_RENDER_SLACK = Buffer.byteLength('… [truncated]', 'utf8')
+
+/**
+ * The shipped {@link CodeRuntime} backend (`ctx.codeRuntime`). Registers as
+ * the `codeRuntime` service; every cap comes from validated config. See the
+ * module doc for the containment model and the class JSDoc on the seam for
+ * the contract this implements (error-as-field, hostile-peer port,
+ * no cross-run state, dispose to quiescence).
+ */
+export class WorkerCodeRuntime extends CodeRuntime {
+  static Config: z<Config> = z.object({
+    computeMs: z.number().default(60_000),
+    maxWallMs: z.number().default(600_000),
+    maxLogBytes: z.number().default(65_536),
+    maxValueBytes: z.number().default(32_768),
+    maxOldGenerationSizeMb: z.number().default(512),
+  })
+
+  readonly language = 'typescript'
+  readonly isolation = 'worker-thread'
+
+  private readonly config: ResolvedConfig
+  private readonly live = new Set<LiveRun>()
+  private disposed = false
+
+  constructor(ctx: Context, config: Config) {
+    super(ctx)
+    // Schemastery filled the defaults; the cast records that. Positivity is a
+    // semantic check the schema's plain number type does not carry.
+    this.config = config as ResolvedConfig
+    for (const [key, value] of Object.entries(this.config)) {
+      if (!(Number.isFinite(value) && value > 0)) throw new Error(`dsh-code-runtime-worker: config.${key} must be a positive number, got ${String(value)}`)
+    }
+    ctx.effect(() => () => this.teardown(), 'worker code-runtime teardown')
+  }
+
+  /**
+   * Dispose to quiescence: mark the service unusable, fail every in-flight
+   * run as aborted, and AWAIT each worker's exit so no worker outlives the
+   * fiber.
+   */
+  private async teardown(): Promise<void> {
+    this.disposed = true
+    const runs = [...this.live]
+    for (const run of runs) run.settle({ kind: 'abort', message: 'runtime disposed' })
+    await Promise.all(runs.map(run => run.finished))
+  }
+
+  /**
+   * Execute one program in a fresh worker. Program outcomes — including a
+   * type-strip syntax error, which never spawns a worker — resolve with
+   * `result.error`; the method rejects only for seam misuse (a disposed
+   * runtime, an invalid binding namespace).
+   * @param request - the program, its bindings, and the abort signal.
+   * @returns the run's outcome per the seam contract.
+   */
+  async run(request: CodeRunRequest): Promise<CodeRunResult> {
+    if (this.disposed) throw new Error('dsh-code-runtime-worker: run() after disposal')
+    const bindings = this.validateBindings(request)
+    if (request.signal?.aborted) {
+      return { logs: [], error: { kind: 'abort', message: String(request.signal.reason) } }
+    }
+
+    let code: string
+    try {
+      const stripped = stripTypeScriptTypes(STRIP_WRAP.prefix + request.program + STRIP_WRAP.suffix)
+      code = stripped.slice(STRIP_WRAP.prefix.length, stripped.length - STRIP_WRAP.suffix.length)
+    } catch (error: unknown) {
+      // A program that does not survive the type-strip (syntax error,
+      // non-erasable syntax like `enum`) is a program failure, reported the
+      // same way a thrown exception would be — and no worker ever spawns.
+      return { logs: [], error: { kind: 'exception', message: messageOf(error) } }
+    }
+
+    return await this.execute(request, code, bindings)
+  }
+
+  /** Reject (seam misuse) malformed binding namespaces: non-identifier or reserved globals, duplicates, and the `console` collision. */
+  private validateBindings(request: CodeRunRequest): Map<string, Record<string, CodeBindingFunction>> {
+    const bindings = new Map<string, Record<string, CodeBindingFunction>>()
+    for (const namespace of request.bindings) {
+      if (!IDENTIFIER.test(namespace.global) || RESERVED_WORDS.has(namespace.global)) {
+        throw new Error(`dsh-code-runtime-worker: binding global ${JSON.stringify(namespace.global)} is not a usable identifier`)
+      }
+      if (namespace.global === 'console' || bindings.has(namespace.global)) {
+        throw new Error(`dsh-code-runtime-worker: duplicate binding global ${JSON.stringify(namespace.global)}`)
+      }
+      bindings.set(namespace.global, namespace.functions)
+    }
+    return bindings
+  }
+
+  /** Spawn the worker for one validated, type-stripped run and drive it to settlement. */
+  private execute(
+    request: CodeRunRequest,
+    code: string,
+    bindings: Map<string, Record<string, CodeBindingFunction>>,
+  ): Promise<CodeRunResult> {
+    const bootData: WorkerBootData = {
+      code,
+      namespaces: [...bindings].map(([global, functions]) => ({ global, names: Object.keys(functions) })),
+      maxLogBytes: this.config.maxLogBytes,
+      maxValueBytes: this.config.maxValueBytes,
+    }
+    const worker = new Worker(WORKER_URL, {
+      workerData: bootData,
+      // Model code gets NO ambient environment — stronger than the scrubbed
+      // env the defensive-patterns rule requires for spawned commands.
+      env: {},
+      // Hermetic flags too: without this the worker inherits the host
+      // process's execArgv (a test runner's or tsx's loader hooks), which a
+      // bare isolate with an empty environment cannot satisfy. The entry
+      // needs nothing beyond native type stripping, on this repo's whole
+      // Node range.
+      execArgv: [],
+      resourceLimits: { maxOldGenerationSizeMb: this.config.maxOldGenerationSizeMb },
+      // Backstop capture: the bootstrap patches JS-level writes into its own
+      // ordered buffer, so these pipes normally stay silent; anything that
+      // still arrives (native-level writes) is appended after the done logs.
+      stdout: true,
+      stderr: true,
+    })
+
+    return new Promise<CodeRunResult>((resolve) => {
+      let settled = false
+      const answered = new Set<number>()
+      const logs: CodeLogEntry[] = []
+      const strayLogs: CodeLogEntry[] = []
+
+      // ONE host-side ledger for everything that lands in `logs`/`strayLogs`,
+      // whatever the path: honest port entries, FORGED port entries (model
+      // code posting `log` messages directly, bypassing the worker-side
+      // LogBuffer), and stray pipe bytes. On the first overflow it emits the
+      // same in-band marker the worker's LogBuffer would and drops the rest,
+      // so the documented cap is one shared `maxLogBytes` however it is hit.
+      let logBudget = this.config.maxLogBytes
+      let logsTruncated = false
+      const admit = (entry: CodeLogEntry, sink: CodeLogEntry[]): void => {
+        if (logsTruncated) return
+        const cost = Buffer.byteLength(entry.text, 'utf8')
+        if (cost > logBudget) {
+          logsTruncated = true
+          sink.push({ source: 'stderr', text: logTruncationMarker(this.config.maxLogBytes) })
+          return
+        }
+        logBudget -= cost
+        sink.push(entry)
+      }
+
+      // No settled guard: `finish` snapshots the arrays when it resolves, so
+      // a chunk flushing after settlement mutates only the discarded buffers,
+      // and the ledger bounds that growth until the pipes close.
+      const captureStray = (source: 'stdout' | 'stderr') => (chunk: Buffer) => {
+        admit({ source, text: chunk.toString('utf8') }, strayLogs)
+      }
+      worker.stdout.on('data', captureStray('stdout'))
+      worker.stderr.on('data', captureStray('stderr'))
+
+      // Settlement: exactly one outcome wins; every path funnels through
+      // here, cleans up the timers/listeners, terminates the worker, and
+      // resolves only after the worker actually exited (quiescence). Logs
+      // streamed eagerly before the settlement are kept — a timed-out or
+      // killed program still shows the model what it printed.
+      let finishResolve!: () => void
+      const finished = new Promise<void>((done) => { finishResolve = done })
+      const finish = (result: Omit<CodeRunResult, 'logs'>): void => {
+        if (settled) return
+        settled = true
+        clearInterval(eluTimer)
+        clearTimeout(wallTimer)
+        request.signal?.removeEventListener('abort', onAbort)
+        this.live.delete(live)
+        void worker.terminate().then(() => {
+          finishResolve()
+          resolve({ ...result, logs: [...logs, ...strayLogs] })
+        })
+      }
+
+      const onDone = (message: WorkerToHost): void => {
+        if (message.type !== 'done') return
+        // Re-cap the completion value HOST-side: the honest path already
+        // capped it in the worker (prepareValue there), but a forged done
+        // message bypasses the bootstrap entirely — without this, model code
+        // could flood the host past maxValueBytes. Honest values pass
+        // unchanged (see VALUE_RENDER_SLACK); the error text is bounded too.
+        finish({
+          ...prepareValue(message.value, this.config.maxValueBytes + VALUE_RENDER_SLACK),
+          ...message.error ? { error: { kind: 'exception' as const, message: truncateUtf8Bytes(message.error.message, this.config.maxValueBytes) } } : {},
+        })
+      }
+
+      const onCall = (message: WorkerToHost): void => {
+        if (message.type !== 'call' || settled) return
+        // Hostile-peer rules: a duplicate id is ignored, an unknown name is
+        // answered with a failure, and a binding throw/reject becomes the
+        // program-side rejection — contained here, never a host crash.
+        if (answered.has(message.id)) return
+        answered.add(message.id)
+        const reply = (payload: ReplyMessage): void => {
+          if (settled) return
+          try {
+            worker.postMessage(payload)
+          } catch {
+            // The reply value failed structured clone; renegotiate as an error
+            // reply, which is always clone-plain. Nothing else throws here.
+            worker.postMessage({ type: 'reply', id: message.id, ok: false, message: 'binding resolution is not structured-cloneable' })
+          }
+        }
+        const record = bindings.get(message.global)
+        // Own-property lookup only: a forged name like 'constructor' or
+        // 'hasOwnProperty' must not walk the record's prototype chain and
+        // reach a callable the consumer never declared.
+        const fn = record && Object.hasOwn(record, message.name) ? record[message.name] : undefined
+        if (typeof fn !== 'function') {
+          reply({ type: 'reply', id: message.id, ok: false, message: `unknown binding ${JSON.stringify(`${message.global}.${message.name}`)}` })
+          return
+        }
+        void (async () => {
+          try {
+            reply({ type: 'reply', id: message.id, ok: true, value: await fn(message.args) })
+          } catch (error: unknown) {
+            reply({ type: 'reply', id: message.id, ok: false, message: messageOf(error) })
+          }
+        })()
+      }
+
+      worker.on('message', (raw: unknown) => {
+        // Parse before touching: the peer can post ANY shape, and a throw in
+        // this listener would crash the host process. Junk drops silently.
+        const message = parseWorkerMessage(raw)
+        if (!message) return
+        if (message.type === 'log' && !settled) admit(message.entry, logs)
+        onCall(message)
+        onDone(message)
+      })
+      worker.on('error', (error: Error) => {
+        finish({ error: { kind: 'worker-exit', message: `worker error: ${error.message}` } })
+      })
+      worker.on('exit', (exitCode: number) => {
+        finish({ error: { kind: 'worker-exit', message: `worker exited with code ${exitCode} before completing` } })
+      })
+
+      // The compute budget reads the worker's own measured busy time, so a
+      // hot loop expires it no matter what dispatches are in flight, while a
+      // program idling on a slow binding accrues nothing.
+      const eluTimer = setInterval(() => {
+        const elu = worker.performance.eventLoopUtilization()
+        if (elu.active > this.config.computeMs) {
+          finish({ error: { kind: 'timeout', message: `compute budget exhausted (${this.config.computeMs}ms busy)` } })
+        }
+      }, ELU_POLL_INTERVAL_MS)
+      const wallTimer = setTimeout(() => {
+        finish({ error: { kind: 'timeout', message: `wall-clock ceiling reached (${this.config.maxWallMs}ms)` } })
+      }, this.config.maxWallMs)
+      const onAbort = (): void => {
+        finish({ error: { kind: 'abort', message: String(request.signal?.reason) } })
+      }
+      request.signal?.addEventListener('abort', onAbort, { once: true })
+
+      const live: LiveRun = {
+        worker,
+        finished,
+        settle: (failure: CodeRunFailure) => { finish({ error: failure }) },
+      }
+      this.live.add(live)
+    })
+  }
+}
+
+export default WorkerCodeRuntime

packages/code-runtime/code-runtime-worker/src/protocol.ts

@@ -0,0 +1,78 @@
+/**
+ * Wire protocol between the host runtime and the worker bootstrap. Everything
+ * crossing the message port is structured-clone-plain and versionless — both
+ * ends ship in this package, always at the same version. The host treats
+ * inbound traffic as HOSTILE (the worker runs model code, which can reach
+ * `parentPort` via `import('node:worker_threads')` and forge any of these
+ * shapes); the worker treats inbound traffic as trusted.
+ *
+ * @module @deepseek-ai/dsh-code-runtime-worker/src/protocol
+ */
+
+import type { CodeLogEntry } from '@deepseek-ai/dsh-code-runtime'
+
+/** What the host hands the worker at spawn, via `workerData`. */
+export interface WorkerBootData {
+  /** The type-stripped (plain JS) program body. */
+  code: string
+  /** Binding namespaces to materialize: the global name plus the function names (functions themselves stay host-side). */
+  namespaces: { global: string; names: string[] }[]
+  /** Shared byte budget for captured log text; exceeding it drops further entries after one in-band marker. */
+  maxLogBytes: number
+  /** Byte cap for the rendered completion value (see the value-preparation contract in bootstrap.ts). */
+  maxValueBytes: number
+}
+
+/** Worker → host: one bridged binding call. */
+export interface CallMessage {
+  type: 'call'
+  /** Worker-issued correlation id; the host answers each id at most once and ignores duplicates. */
+  id: number
+  /** The namespace global the call targets. */
+  global: string
+  /** The function name within the namespace. */
+  name: string
+  /** The single argument, structured-clone-plain. */
+  args: unknown
+}
+
+/** Worker → host: one captured log entry, streamed eagerly so output survives a mid-run termination (timeout, abort, OOM). */
+export interface LogMessage {
+  type: 'log'
+  entry: CodeLogEntry
+}
+
+/**
+ * Worker → host: the program settled. `error` carries a program exception
+ * (the only failure the bootstrap itself can report — budgets, aborts, and
+ * substrate death are observed host-side). `value` is present only on a
+ * clean completion that produced one (already size-capped and
+ * clone-safe per the bootstrap's value preparation). Logs are NOT carried
+ * here — they streamed eagerly as {@link LogMessage}s.
+ */
+export interface DoneMessage {
+  type: 'done'
+  value?: unknown
+  error?: { message: string }
+}
+
+/** Every message the worker sends. */
+export type WorkerToHost = CallMessage | LogMessage | DoneMessage
+
+/** Host → worker: the answer to one {@link CallMessage}. */
+export type ReplyMessage =
+  | { type: 'reply'; id: number; ok: true; value: unknown }
+  | { type: 'reply'; id: number; ok: false; message: string }
+
+/**
+ * The in-band marker entry text announcing that log capture stopped at the
+ * byte budget. Shared wire vocabulary: the worker's LogBuffer emits it when
+ * ITS budget exhausts, and the host emits the identical text when its own
+ * ledger drops an entry first (forged port traffic, stray pipe bytes) — so
+ * a truncated run reads the same however the cap was hit.
+ * @param maxBytes - the configured `maxLogBytes` the marker names.
+ * @returns the marker line.
+ */
+export function logTruncationMarker(maxBytes: number): string {
+  return `[dsh-code-runtime-worker] log capture truncated at ${maxBytes} bytes`
+}

packages/code-runtime/code-runtime-worker/src/worker.ts

@@ -0,0 +1,20 @@
+/**
+ * The worker-thread entrypoint: self-executing glue over
+ * `bootstrap.ts`'s {@link runWorkerMain}, kept to the spawn wiring alone.
+ * Like `bin.ts` CLI entrypoints, this file executes only inside a spawned
+ * worker isolate — a place the coverage provider cannot observe — so it is
+ * excluded from the coverage gate while every line of actual logic lives in
+ * `bootstrap.ts`, unit-tested in-process; the real spawn path is pinned by
+ * the integration tests that run genuine workers.
+ *
+ * @module @deepseek-ai/dsh-code-runtime-worker/src/worker
+ */
+
+import { parentPort, workerData } from 'node:worker_threads'
+import { runWorkerMain } from './bootstrap.ts'
+import type { WorkerBootData } from './protocol.ts'
+
+// A worker always has a parent port; guard loudly rather than run detached.
+if (!parentPort) throw new Error('dsh-code-runtime-worker: worker entry loaded outside a worker thread')
+
+await runWorkerMain(parentPort, workerData as WorkerBootData, { stdout: process.stdout, stderr: process.stderr })

packages/code-runtime/code-runtime-worker/tests/bootstrap.spec.ts

@@ -0,0 +1,273 @@
+import { describe, expect, it } from 'vitest'
+import { EventEmitter } from 'node:events'
+import { LogBuffer, makeConsoleShim, makeNamespaces, captureStreamWrites, prepareValue, runWorkerMain, truncateUtf8Bytes, wireReplies } from '@deepseek-ai/dsh-code-runtime-worker/src/bootstrap.ts'
+import type { BootstrapPort, PatchableStream, PendingCall } from '@deepseek-ai/dsh-code-runtime-worker/src/bootstrap.ts'
+import type { ReplyMessage, WorkerToHost } from '@deepseek-ai/dsh-code-runtime-worker/src/protocol.ts'
+import type { CodeLogEntry } from '@deepseek-ai/dsh-code-runtime'
+
+/**
+ * An in-process stand-in for the worker's parentPort: the test plays the
+ * HOST side — inspect what the bootstrap posted, feed replies back — so
+ * every line of worker-side logic runs under coverage without spawning an
+ * isolate (real-worker behavior is pinned by runtime.spec.ts).
+ */
+class FakePort implements BootstrapPort {
+  sent: WorkerToHost[] = []
+  private readonly emitter = new EventEmitter()
+  /** Host-scripted responder; return undefined to leave the call pending. */
+  respond: (message: WorkerToHost) => ReplyMessage | undefined = () => undefined
+
+  postMessage(message: WorkerToHost): void {
+    this.sent.push(message)
+    const reply = this.respond(message)
+    if (reply) queueMicrotask(() => this.emitter.emit('message', reply))
+  }
+
+  on(event: 'message', listener: (message: ReplyMessage) => void): void {
+    this.emitter.on(event, listener)
+  }
+
+  deliver(message: ReplyMessage): void {
+    this.emitter.emit('message', message)
+  }
+
+  logs(): CodeLogEntry[] {
+    return this.sent.filter(message => message.type === 'log').map(message => message.entry)
+  }
+
+  done(): WorkerToHost | undefined {
+    return this.sent.find(message => message.type === 'done')
+  }
+}
+
+function fakeStreams(): { stdout: PatchableStream; stderr: PatchableStream } {
+  return { stdout: { write: () => true }, stderr: { write: () => true } }
+}
+
+const BOOT = { maxLogBytes: 65_536, maxValueBytes: 32_768 }
+
+describe('LogBuffer', () => {
+  it('streams entries to the sink until the byte budget, then emits one marker and drops the rest', () => {
+    const seen: CodeLogEntry[] = []
+    const buffer = new LogBuffer(10, entry => seen.push(entry))
+    buffer.push({ source: 'console', level: 'log', text: '12345' })
+    buffer.push({ source: 'console', level: 'log', text: '123456' })
+    buffer.push({ source: 'console', level: 'log', text: 'dropped' })
+    expect(seen.map(entry => entry.text)).toEqual([
+      '12345',
+      '[dsh-code-runtime-worker] log capture truncated at 10 bytes',
+    ])
+  })
+})
+
+describe('makeConsoleShim', () => {
+  it('captures the five levels and renders non-strings inspect-style', () => {
+    const seen: CodeLogEntry[] = []
+    const shim = makeConsoleShim(new LogBuffer(1_000, entry => seen.push(entry)))
+    shim.log('plain', { a: 1 })
+    shim.info('i')
+    shim.warn('w')
+    shim.error('e')
+    shim.debug('d')
+    expect(seen.map(entry => entry.level)).toEqual(['log', 'info', 'warn', 'error', 'debug'])
+    expect(seen[0]?.text).toBe('plain { a: 1 }')
+    expect(seen.every(entry => entry.source === 'console')).toBe(true)
+  })
+})
+
+describe('captureStreamWrites', () => {
+  it('redirects writes into the buffer and restores on request', () => {
+    const seen: CodeLogEntry[] = []
+    const buffer = new LogBuffer(1_000, entry => seen.push(entry))
+    let underlying = ''
+    const stream: PatchableStream = { write: (chunk: unknown) => { underlying += String(chunk); return true } }
+    const restore = captureStreamWrites(buffer, stream, 'stdout')
+    stream.write('captured', 'utf8')
+    stream.write(Buffer.from('bytes'))
+    restore()
+    stream.write('after')
+    expect(seen.map(entry => entry.text)).toEqual(['captured', 'bytes'])
+    expect(seen[0]).toMatchObject({ source: 'stdout' })
+    expect(underlying).toBe('after')
+  })
+
+  it('invokes the write callback asynchronously, in both optional-encoding shapes', async () => {
+    const buffer = new LogBuffer(1_000, () => {})
+    const stream: PatchableStream = { write: () => true }
+    captureStreamWrites(buffer, stream, 'stdout')
+    const calls: (Error | null | undefined)[] = []
+    stream.write('two-arg', (error?: Error | null) => calls.push(error))
+    stream.write('three-arg', 'utf8', (error?: Error | null) => calls.push(error))
+    // Node's contract: the callback fires after the write call returns.
+    expect(calls).toEqual([])
+    await new Promise<void>(resolve => stream.write('awaited flush', resolve))
+    expect(calls).toEqual([null, null])
+  })
+
+  it('still fires the callback for a write the exhausted budget drops', async () => {
+    const buffer = new LogBuffer(4, () => {})
+    const stream: PatchableStream = { write: () => true }
+    captureStreamWrites(buffer, stream, 'stdout')
+    stream.write('this write overflows the budget and is dropped')
+    await new Promise<void>(resolve => stream.write('also dropped', resolve))
+  })
+})
+
+describe('prepareValue', () => {
+  it('omits undefined, passes small cloneable values raw', () => {
+    expect(prepareValue(undefined, 100)).toEqual({})
+    expect(prepareValue({ a: [1, 'two'] }, 100)).toEqual({ value: { a: [1, 'two'] } })
+  })
+
+  it('replaces a non-cloneable value with its rendering', () => {
+    const { value } = prepareValue({ fn: () => 1 }, 1_000)
+    expect(typeof value).toBe('string')
+    expect(value).toContain('fn')
+  })
+
+  it('replaces an oversized value with a truncation-marked capped rendering', () => {
+    const { value } = prepareValue('x'.repeat(50), 10)
+    expect(value).toBe(`${'x'.repeat(10)}… [truncated]`)
+  })
+
+  it('measures a container by its structured-clone wire size, not its bounded rendering', () => {
+    // The bounded inspect rendering of a huge array is tiny ("... N more
+    // items"), but its real cross-boundary size is not — the cap must catch
+    // it, replacing the value with that bounded rendering.
+    const huge = new Array(50_000).fill(7)
+    const { value } = prepareValue(huge, 1_000)
+    expect(typeof value).toBe('string')
+    expect(value).toContain('more items')
+  })
+
+  it('caps a multibyte string by UTF-8 bytes, not UTF-16 length', () => {
+    // 4 code units but 12 UTF-8 bytes: a length-counting cap would pass the
+    // full string through untruncated.
+    expect(prepareValue('€€€€', 4)).toEqual({ value: '€… [truncated]' })
+  })
+
+  it('caps a multibyte rendering by UTF-8 bytes too', () => {
+    // Wire size (24-byte string inside an array) exceeds the cap, so the
+    // value crosses as its rendering — whose truncation must also be
+    // byte-exact: "[ '" (3 bytes) + two € (6 bytes) = 9; a third € would
+    // overflow the 10-byte budget.
+    expect(prepareValue(['€€€€€€€€'], 10)).toEqual({ value: "[ '€€… [truncated]" })
+  })
+})
+
+describe('truncateUtf8Bytes', () => {
+  it('returns a fitting string whole', () => {
+    expect(truncateUtf8Bytes('fits', 4)).toBe('fits')
+  })
+
+  it('cuts at a code-point boundary, never mid-surrogate-pair', () => {
+    // Each 😀 is one code point, two code units, four UTF-8 bytes: a 5-byte
+    // budget fits exactly one — and never leaves a lone surrogate behind.
+    const cut = truncateUtf8Bytes('😀😀', 5)
+    expect(cut).toBe('😀')
+    expect(Buffer.byteLength(truncateUtf8Bytes('😀😀', 3), 'utf8')).toBe(0)
+  })
+})
+
+describe('makeNamespaces', () => {
+  it('exposes prototype-colliding names as ordinary own properties', async () => {
+    const port = new FakePort()
+    port.respond = message => message.type === 'call' ? { type: 'reply', id: message.id, ok: true, value: `${message.name}-ok` } : undefined
+    const pending = new Map<number, PendingCall>()
+    wireReplies(port, pending)
+    const [tools] = makeNamespaces({ namespaces: [{ global: 'tools', names: ['__proto__', 'constructor', 'toString'] }] }, port, pending, { value: 1 }) as [Record<string, (args: unknown) => Promise<unknown>>]
+    expect(Object.getPrototypeOf(tools)).toBeNull()
+    await expect(tools['__proto__']?.({})).resolves.toBe('__proto__-ok')
+    await expect(tools['constructor']?.({})).resolves.toBe('constructor-ok')
+    await expect(tools['toString']?.({})).resolves.toBe('toString-ok')
+  })
+
+  it('rejects a non-cloneable argument without leaking the pending entry', async () => {
+    let firstCall = true
+    const throwingPort: BootstrapPort = {
+      // First call throws an Error (the real DataCloneError shape), the
+      // second a bare string — the rejection renders both.
+      postMessage: () => {
+        if (firstCall) { firstCall = false; throw new Error('DataCloneError-ish') }
+        throw 'raw-clone-failure'
+      },
+      on: () => {},
+    }
+    const pending = new Map<number, PendingCall>()
+    const [tools] = makeNamespaces({ namespaces: [{ global: 'tools', names: ['x'] }] }, throwingPort, pending, { value: 1 }) as [Record<string, (args: unknown) => Promise<unknown>>]
+    await expect(tools.x?.(() => 1)).rejects.toThrow(/structured-cloneable: DataCloneError-ish/)
+    await expect(tools.x?.(() => 1)).rejects.toThrow(/structured-cloneable: raw-clone-failure/)
+    expect(pending.size).toBe(0)
+  })
+})
+
+describe('runWorkerMain', () => {
+  it('runs a program end-to-end: bindings, console, return value', async () => {
+    const port = new FakePort()
+    port.respond = message => message.type === 'call' ? { type: 'reply', id: message.id, ok: true, value: (message.args as { n: number }).n * 2 } : undefined
+    await runWorkerMain(port, {
+      ...BOOT,
+      code: 'const doubled = await tools.double({ n: 21 }); console.log("got", doubled); return { doubled };',
+      namespaces: [{ global: 'tools', names: ['double'] }],
+    }, fakeStreams())
+    expect(port.logs()).toEqual([{ source: 'console', level: 'log', text: 'got 42' }])
+    expect(port.done()).toEqual({ type: 'done', value: { doubled: 42 } })
+  })
+
+  it('reports a thrown program error on the done message', async () => {
+    const port = new FakePort()
+    await runWorkerMain(port, { ...BOOT, code: 'throw new Error("boom")', namespaces: [] }, fakeStreams())
+    const done = port.done()
+    expect(done?.type).toBe('done')
+    expect(done?.type === 'done' ? done.error?.message : undefined).toContain('boom')
+    expect(done?.type === 'done' ? done.value : undefined).toBeUndefined()
+  })
+
+  it('renders non-Error throws and stack-less Errors on the done message', async () => {
+    const rawPort = new FakePort()
+    await runWorkerMain(rawPort, { ...BOOT, code: 'throw "raw-throw"', namespaces: [] }, fakeStreams())
+    expect(rawPort.done()).toEqual({ type: 'done', error: { message: 'raw-throw' } })
+
+    const barePort = new FakePort()
+    await runWorkerMain(barePort, { ...BOOT, code: 'const e = new Error("bare"); e.stack = undefined; throw e', namespaces: [] }, fakeStreams())
+    expect(barePort.done()).toEqual({ type: 'done', error: { message: 'bare' } })
+  })
+
+  it('surfaces a host failure reply as a program-side rejection it can catch', async () => {
+    const port = new FakePort()
+    port.respond = message => message.type === 'call' ? { type: 'reply', id: message.id, ok: false, message: 'denied by host' } : undefined
+    await runWorkerMain(port, {
+      ...BOOT,
+      code: 'try { await tools.x({}) } catch (error) { return `caught: ${error.message}` }',
+      namespaces: [{ global: 'tools', names: ['x'] }],
+    }, fakeStreams())
+    expect(port.done()).toEqual({ type: 'done', value: 'caught: denied by host' })
+  })
+
+  it('ignores replies for unknown pending ids', async () => {
+    const port = new FakePort()
+    port.respond = (message) => {
+      if (message.type !== 'call') return undefined
+      // Deliver a stray reply first; the real one follows.
+      port.deliver({ type: 'reply', id: 9_999, ok: true, value: 'stray' })
+      return { type: 'reply', id: message.id, ok: true, value: 'real' }
+    }
+    await runWorkerMain(port, {
+      ...BOOT,
+      code: 'return await tools.x({})',
+      namespaces: [{ global: 'tools', names: ['x'] }],
+    }, fakeStreams())
+    expect(port.done()).toEqual({ type: 'done', value: 'real' })
+  })
+
+  it('captures raw stream writes through the patched process streams', async () => {
+    const port = new FakePort()
+    const streams = fakeStreams()
+    await runWorkerMain(port, { ...BOOT, code: 'return 1', namespaces: [] }, streams)
+    streams.stdout.write('never seen — already restored? no: patch persists in worker')
+    // The patch stays installed for the worker's lifetime; writes during the
+    // program landed in order. Here the program wrote nothing via streams, so
+    // only the post-run write above went through the patched slot.
+    expect(port.logs().at(-1)).toMatchObject({ source: 'stdout' })
+  })
+})

packages/code-runtime/code-runtime-worker/tests/built-lib.e2e.ts

@@ -0,0 +1,55 @@
+import { spawn } from 'node:child_process'
+import { existsSync } from 'node:fs'
+import { join } from 'node:path'
+import { fileURLToPath } from 'node:url'
+import { describe, expect, it } from 'vitest'
+
+/**
+ * BUILT-ARTIFACT smoke for the published package (the real-load-path guard
+ * from docs/testing.md): the unit suite runs `src/` under vitest, where the
+ * worker entry resolves to `src/worker.ts` — a consumer runs `lib/index.js`
+ * under plain `node`, where it must resolve the sibling `lib/worker.js`
+ * bundle instead. This spawns plain `node` (NOT tsx) from inside the package
+ * directory and imports the package BY NAME, so resolution flows through the
+ * real `exports` map exactly as it would from a downstream install; the
+ * program exercises the type-strip, the worker spawn, the binding bridge,
+ * and log capture end-to-end through the built bundles.
+ *
+ * It build-gates: SKIPS when the built artifacts are absent (suite run
+ * without `pnpm run build`); CI runs it after the build step. KEYLESS — no
+ * model is involved.
+ */
+
+const pkgDir = fileURLToPath(new URL('..', import.meta.url))
+const built = ['lib/index.js', 'lib/worker.js'].every(file => existsSync(join(pkgDir, file)))
+  && existsSync(join(pkgDir, '../code-runtime/lib/index.js'))
+
+describe.skipIf(!built)('built lib real load path (plain node)', () => {
+  it('runs a TypeScript program with a binding through lib/index.js and its lib/worker.js entry', async () => {
+    const script = `
+      const { Context } = await import('cordis')
+      const { WorkerCodeRuntime } = await import('@deepseek-ai/dsh-code-runtime-worker')
+      const ctx = new Context()
+      await ctx.plugin(WorkerCodeRuntime, {})
+      const result = await ctx.codeRuntime.run({
+        program: 'const doubled: number = await tools.double({ n: 21 }); console.log("halfway", doubled); return doubled;',
+        bindings: [{ global: 'tools', functions: { double: async args => args.n * 2 } }],
+      })
+      console.log(JSON.stringify(result))
+      process.exit(0)
+    `
+    const child = spawn(process.execPath, ['--input-type=module', '-e', script], { cwd: pkgDir, stdio: ['ignore', 'pipe', 'pipe'] })
+    let stdout = ''
+    let stderr = ''
+    child.stdout.on('data', (chunk: Buffer) => { stdout += chunk.toString('utf8') })
+    child.stderr.on('data', (chunk: Buffer) => { stderr += chunk.toString('utf8') })
+    const exitCode = await new Promise<number | null>(resolve => child.on('close', resolve))
+
+    expect(exitCode, `stderr:\n${stderr}`).toBe(0)
+    const lastLine = stdout.trim().split('\n').at(-1) ?? ''
+    const result = JSON.parse(lastLine) as { value?: unknown; logs: { source: string; level?: string; text: string }[]; error?: unknown }
+    expect(result.error).toBeUndefined()
+    expect(result.value).toBe(42)
+    expect(result.logs).toContainEqual({ source: 'console', level: 'log', text: 'halfway 42' })
+  })
+})

packages/code-runtime/code-runtime-worker/tests/runtime.spec.ts

@@ -0,0 +1,451 @@
+import { describe, expect, it } from 'vitest'
+import { Context } from 'cordis'
+import { WorkerCodeRuntime } from '@deepseek-ai/dsh-code-runtime-worker'
+import type { Config } from '@deepseek-ai/dsh-code-runtime-worker'
+import type { CodeRunResult } from '@deepseek-ai/dsh-code-runtime'
+
+/**
+ * Integration suite over REAL worker threads (no mocks — workers are cheap
+ * and local, per docs/testing.md's real-over-mock policy). Each test builds
+ * a fresh context so budgets can be tuned per case.
+ */
+async function setup(config: Config = {}) {
+  const ctx = new Context()
+  await ctx.plugin(WorkerCodeRuntime, config)
+  const runtime = ctx.codeRuntime as WorkerCodeRuntime
+  return { ctx, runtime }
+}
+
+/** Convenience: one namespace `tools` with the given functions. */
+function tools(functions: Record<string, (args: unknown) => Promise<unknown>>) {
+  return [{ global: 'tools', functions }]
+}
+
+describe('WorkerCodeRuntime — programs and bindings (real workers)', () => {
+  it('registers with the seam descriptors', async () => {
+    const { runtime } = await setup()
+    expect(runtime.language).toBe('typescript')
+    expect(runtime.isolation).toBe('worker-thread')
+  })
+
+  it('runs TypeScript (erasable syntax), captures console/stdout in order, returns the value', async () => {
+    const { runtime } = await setup()
+    const result = await runtime.run({
+      program: `
+        interface Point { x: number; y: number }
+        const p: Point = { x: 1, y: 2 } as Point;
+        console.log('point', p);
+        process.stdout.write('raw-out\\n');
+        console.warn('careful');
+        return p.x + p.y;
+      `,
+      bindings: [],
+    })
+    expect(result.error).toBeUndefined()
+    expect(result.value).toBe(3)
+    expect(result.logs.map(entry => [entry.source, entry.level ?? null])).toEqual([
+      ['console', 'log'],
+      ['stdout', null],
+      ['console', 'warn'],
+    ])
+    expect(result.logs[0]?.text).toBe('point { x: 1, y: 2 }')
+  })
+
+  it('bridges binding calls both ways and rejects the program-side call on a host rejection', async () => {
+    const { runtime } = await setup()
+    const calls: unknown[] = []
+    const result = await runtime.run({
+      program: `
+        const first = await tools.echo({ n: 1 });
+        let caught = '';
+        try { await tools.fail({}) } catch (error) { caught = error.message }
+        let caughtRaw = '';
+        try { await tools.failRaw({}) } catch (error) { caughtRaw = error.message }
+        return { first, caught, caughtRaw };
+      `,
+      bindings: tools({
+        echo: async (args) => { calls.push(args); return { echoed: args } },
+        fail: async () => { throw new Error('nope') },
+        // A non-Error throw: the host renders it, the program still catches.
+        failRaw: async () => { throw 'raw-nope' },
+      }),
+    })
+    expect(result.error).toBeUndefined()
+    expect(result.value).toEqual({ first: { echoed: { n: 1 } }, caught: 'nope', caughtRaw: 'raw-nope' })
+    expect(calls).toEqual([{ n: 1 }])
+  })
+
+  it('reports non-erasable syntax as an exception without spawning a worker', async () => {
+    const { runtime } = await setup()
+    const result = await runtime.run({ program: 'enum E { A }\nreturn 1', bindings: [] })
+    expect(result.error?.kind).toBe('exception')
+    expect(result.error?.message).toMatch(/enum|strip/i)
+  })
+
+  it('reports a runtime throw as an exception with the message', async () => {
+    const { runtime } = await setup()
+    const result = await runtime.run({ program: 'throw new Error("kaboom")', bindings: [] })
+    expect(result.error?.kind).toBe('exception')
+    expect(result.error?.message).toContain('kaboom')
+  })
+
+  it('gives the program an EMPTY environment', async () => {
+    const { runtime } = await setup()
+    const result = await runtime.run({ program: 'return JSON.stringify(process.env)', bindings: [] })
+    expect(result.value).toBe('{}')
+  })
+
+  it('replaces a non-cloneable return value with a string rendering', async () => {
+    const { runtime } = await setup()
+    const result = await runtime.run({ program: 'return { f: () => 1 }', bindings: [] })
+    expect(typeof result.value).toBe('string')
+  })
+
+  it('completes a program that returns nothing with no value at all', async () => {
+    const { runtime } = await setup()
+    const result = await runtime.run({ program: 'const x = 1', bindings: [] })
+    expect(result.error).toBeUndefined()
+    expect('value' in result).toBe(false)
+  })
+
+  it('keeps logs streamed before a failure', async () => {
+    const { runtime } = await setup()
+    const result = await runtime.run({
+      program: 'console.log("before"); throw new Error("after-log")',
+      bindings: [],
+    })
+    expect(result.error?.kind).toBe('exception')
+    expect(result.logs.map(entry => entry.text)).toContain('before')
+  })
+})
+
+describe('WorkerCodeRuntime — budgets and containment (real workers)', () => {
+  it('ends a hot loop at the compute budget — including behind a pending decoy dispatch', async () => {
+    const { runtime } = await setup({ computeMs: 300, maxWallMs: 30_000 })
+    const result = await runtime.run({
+      // The decoy: fire a call at a never-resolving binding WITHOUT awaiting,
+      // then spin. Host-side pending-call bookkeeping would pause a naive
+      // budget here; measured busy time cannot be fooled.
+      program: 'void tools.slow({}); for (;;) {}',
+      bindings: tools({ slow: () => new Promise(() => {}) }),
+    })
+    expect(result.error?.kind).toBe('timeout')
+    expect(result.error?.message).toContain('compute budget')
+  }, 15_000)
+
+  it('does not charge time spent awaiting a slow binding against the compute budget', async () => {
+    const { runtime } = await setup({ computeMs: 250, maxWallMs: 30_000 })
+    const result = await runtime.run({
+      program: 'return await tools.slow({})',
+      bindings: tools({ slow: () => new Promise(resolve => setTimeout(() => { resolve('slow-done') }, 700)) }),
+    })
+    expect(result.error).toBeUndefined()
+    expect(result.value).toBe('slow-done')
+  }, 15_000)
+
+  it('ends an idle-forever run at the wall-clock ceiling', async () => {
+    const { runtime } = await setup({ computeMs: 30_000, maxWallMs: 400 })
+    const result = await runtime.run({
+      program: 'await tools.never({}); return 1',
+      bindings: tools({ never: () => new Promise(() => {}) }),
+    })
+    expect(result.error?.kind).toBe('timeout')
+    expect(result.error?.message).toContain('wall-clock ceiling')
+  }, 15_000)
+
+  it('reports an abort mid-run and stops the worker', async () => {
+    const { runtime } = await setup()
+    const controller = new AbortController()
+    setTimeout(() => { controller.abort('user-cancel') }, 150)
+    const result = await runtime.run({ program: 'for (;;) {}', bindings: [], signal: controller.signal })
+    expect(result.error).toEqual({ kind: 'abort', message: 'user-cancel' })
+  }, 15_000)
+
+  it('reports a pre-aborted signal without spawning', async () => {
+    const { runtime } = await setup()
+    const controller = new AbortController()
+    controller.abort('too-late')
+    const result = await runtime.run({ program: 'return 1', bindings: [], signal: controller.signal })
+    expect(result.error).toEqual({ kind: 'abort', message: 'too-late' })
+  })
+
+  it('drops a binding resolution that lands after the run settled', async () => {
+    const { runtime } = await setup()
+    const controller = new AbortController()
+    let replyDelivered!: Promise<void>
+    const result = await runtime.run({
+      program: 'void tools.late({}); for (;;) {}',
+      bindings: tools({
+        // Anchored on invocation: abort 100ms after the call reaches the
+        // host, resolve 400ms after — by then the run has settled, so the
+        // resolution's reply hits the post-settlement drop.
+        late: () => new Promise((resolve) => {
+          setTimeout(() => { controller.abort('cancel-now') }, 100)
+          replyDelivered = new Promise(done => setTimeout(() => { resolve('too-late'); done() }, 400))
+        }),
+      }),
+      signal: controller.signal,
+    })
+    expect(result.error).toEqual({ kind: 'abort', message: 'cancel-now' })
+    // Let the late resolution actually fire so its reply executes instead of
+    // being cancelled with the test.
+    await replyDelivered
+  }, 15_000)
+
+  it('contains an OOM under resourceLimits as worker-exit, host process healthy', async () => {
+    const { runtime } = await setup({ maxOldGenerationSizeMb: 32 })
+    const result = await runtime.run({
+      program: 'const hog = []; for (;;) hog.push(new Array(1e6).fill(1));',
+      bindings: [],
+    })
+    expect(result.error?.kind).toBe('worker-exit')
+    // And the host is fine: run something else.
+    const after = await runtime.run({ program: 'return "alive"', bindings: [] })
+    expect(after.value).toBe('alive')
+  }, 30_000)
+
+  it('truncates runaway log output at the byte budget with an in-band marker', async () => {
+    const { runtime } = await setup({ maxLogBytes: 300 })
+    const result = await runtime.run({
+      program: 'for (let i = 0; i < 1000; i++) console.log("spam line", i); return 1',
+      bindings: [],
+    })
+    expect(result.logs.at(-1)?.text).toContain('truncated at 300 bytes')
+    const total = result.logs.reduce((sum, entry) => sum + Buffer.byteLength(entry.text, 'utf8'), 0)
+    expect(total).toBeLessThan(1_000)
+  })
+
+  it('caps an oversized return value with a truncation marker', async () => {
+    const { runtime } = await setup({ maxValueBytes: 64 })
+    const result = await runtime.run({ program: 'return "y".repeat(10_000)', bindings: [] })
+    expect(result.value).toBe(`${'y'.repeat(64)}… [truncated]`)
+  })
+
+  it('caps a multibyte return value by UTF-8 bytes, not string length', async () => {
+    // 4 code units, 12 UTF-8 bytes: a length-counting cap would let the full
+    // string cross. The worker's byte-exact capped rendering then passes the
+    // host re-cap unchanged (cap + marker is exactly the granted slack).
+    const { runtime } = await setup({ maxValueBytes: 4 })
+    const result = await runtime.run({ program: 'return "€€€€"', bindings: [] })
+    expect(result.value).toBe('€… [truncated]')
+  })
+
+  it('completes a program that awaits its write callback, capturing the chunk', async () => {
+    // Node's write(chunk[, encoding][, callback]) contract: dropping the
+    // callback would leave this promise pending until the wall ceiling and
+    // misreport a completed program as a timeout.
+    const { runtime } = await setup({ maxWallMs: 2_000 })
+    const result = await runtime.run({
+      program: 'await new Promise(resolve => process.stdout.write("flushed", resolve)); return "done"',
+      bindings: [],
+    })
+    expect(result.error).toBeUndefined()
+    expect(result.value).toBe('done')
+    expect(result.logs).toContainEqual({ source: 'stdout', text: 'flushed' })
+  })
+
+  it('caps a huge container whose bounded rendering is small (wire size, not rendering, is what counts)', async () => {
+    const { runtime } = await setup()
+    const result = await runtime.run({ program: 'return new Array(50_000).fill(7)', bindings: [] })
+    expect(result.error).toBeUndefined()
+    expect(typeof result.value).toBe('string')
+    expect(result.value).toContain('more items')
+  })
+
+  it('captures pipe writes that bypass the patched write slot as stray logs, capped by the same budget', async () => {
+    const { runtime } = await setup({ maxLogBytes: 4 })
+    const result = await runtime.run({
+      // The bootstrap patches the stream instance's own `write`; going
+      // through the prototype's slot reaches the real pipe underneath, so
+      // the bytes arrive host-side as stray data. The pauses keep the two
+      // writes in separate pipe chunks and let them land before settlement.
+      program: `
+        const write = (text) => Object.getPrototypeOf(process.stdout).write.call(process.stdout, text);
+        write('abcd');
+        await new Promise(resolve => setTimeout(resolve, 150));
+        write('ef');
+        await new Promise(resolve => setTimeout(resolve, 100));
+        return 1;
+      `,
+      bindings: [],
+    })
+    expect(result.error).toBeUndefined()
+    expect(result.logs).toContainEqual({ source: 'stdout', text: 'abcd' })
+    expect(result.logs.map(entry => entry.text)).not.toContain('ef')
+  }, 15_000)
+})
+
+describe('WorkerCodeRuntime — hostile programs (real workers)', () => {
+  it('survives forged port traffic: unknown binding names, duplicate ids, junk shapes', async () => {
+    const { runtime } = await setup()
+    const result = await runtime.run({
+      program: `
+        const { parentPort } = await import('node:worker_threads');
+        parentPort.postMessage({ type: 'call', id: 7777, global: 'tools', name: 'missing', args: {} });
+        parentPort.postMessage({ type: 'call', id: 7777, global: 'tools', name: 'missing', args: {} });
+        parentPort.postMessage({ type: 'call', id: 7778, global: 'tools', name: 'constructor', args: {} });
+        parentPort.postMessage({ type: 'junk' });
+        return await tools.real({});
+      `,
+      bindings: tools({ real: async () => 'still-works' }),
+    })
+    expect(result.error).toBeUndefined()
+    expect(result.value).toBe('still-works')
+  })
+
+  it('survives arbitrary junk on the port: non-objects, junk types, malformed calls, logs, and dones', async () => {
+    const { runtime } = await setup()
+    const result = await runtime.run({
+      program: `
+        const { parentPort } = await import('node:worker_threads');
+        for (const junk of [
+          null, 42, 'junk', [],
+          { type: 'nope' },
+          { type: 'call' },
+          { type: 'call', id: 'x', global: 'tools', name: 'real', args: {} },
+          { type: 'call', id: 1e9, global: 7, name: 'real', args: {} },
+          { type: 'call', id: 1e9, global: 'tools', name: 7, args: {} },
+          { type: 'log' },
+          { type: 'log', entry: null },
+          { type: 'log', entry: { source: 'stdout', text: 7 } },
+          { type: 'log', entry: { source: 'nope', text: 'x' } },
+          { type: 'log', entry: { source: 'console', level: 'nope', text: 'x' } },
+          { type: 'log', entry: { source: 'console', level: 7, text: 'x' } },
+          { type: 'done', error: 5 },
+          { type: 'done', error: { message: 5 } },
+        ]) parentPort.postMessage(junk);
+        return await tools.real({});
+      `,
+      bindings: tools({ real: async () => 'still-works' }),
+    })
+    expect(result.error).toBeUndefined()
+    expect(result.value).toBe('still-works')
+    expect(result.logs).toEqual([])
+  })
+
+  it('caps forged log floods and forged done values at the configured budgets, dropping forged extra fields', async () => {
+    const { runtime } = await setup({ maxLogBytes: 200, maxValueBytes: 64 })
+    const result = await runtime.run({
+      // Forged messages bypass the worker-side LogBuffer and prepareValue
+      // entirely — only the host-side ledger and re-cap stand between model
+      // code and an unbounded result.
+      program: `
+        const { parentPort } = await import('node:worker_threads');
+        for (let i = 0; i < 50; i++) parentPort.postMessage({ type: 'log', entry: { source: 'stdout', text: 'F'.repeat(100), forged: true } });
+        parentPort.postMessage({ type: 'done', value: 'V'.repeat(100000) });
+        for (;;) {}
+      `,
+      bindings: [],
+    })
+    expect(typeof result.value).toBe('string')
+    const value = result.value as string
+    expect(value.startsWith('V'.repeat(64))).toBe(true)
+    expect(value.endsWith('… [truncated]')).toBe(true)
+    expect(value.length).toBeLessThan(120)
+    const marker = '[dsh-code-runtime-worker] log capture truncated at 200 bytes'
+    const total = result.logs.reduce((sum, entry) => sum + Buffer.byteLength(entry.text, 'utf8'), 0)
+    expect(total).toBeLessThanOrEqual(200 + Buffer.byteLength(marker, 'utf8'))
+    expect(result.logs.at(-1)?.text).toBe(marker)
+    expect(result.logs.every(entry => !('forged' in entry))).toBe(true)
+  })
+
+  it('accepts a forged done carrying both value and error (self-sabotage, contained)', async () => {
+    const { runtime } = await setup()
+    const result = await runtime.run({
+      program: `
+        const { parentPort } = await import('node:worker_threads');
+        parentPort.postMessage({ type: 'done', value: 'lied', error: { message: 'fake failure' } });
+        for (;;) {}
+      `,
+      bindings: [],
+    })
+    expect(result.value).toBe('lied')
+    expect(result.error).toEqual({ kind: 'exception', message: 'fake failure' })
+  })
+
+  it('byte-bounds forged multibyte error text at the host', async () => {
+    // Forged error text bypasses the worker entirely; the host bound is a
+    // BYTE bound (two € = 6 bytes fit an 8-byte cap, a third would not).
+    const { runtime } = await setup({ maxValueBytes: 8 })
+    const result = await runtime.run({
+      program: `
+        const { parentPort } = await import('node:worker_threads');
+        parentPort.postMessage({ type: 'done', error: { message: '€'.repeat(1000) } });
+        for (;;) {}
+      `,
+      bindings: [],
+    })
+    expect(result.error).toEqual({ kind: 'exception', message: '€€' })
+  })
+
+  it('answers a binding whose resolution cannot be cloned with a failure reply', async () => {
+    const { runtime } = await setup()
+    const result = await runtime.run({
+      program: 'try { await tools.bad({}) } catch (error) { return error.message }',
+      bindings: tools({ bad: async () => (() => 1) }),
+    })
+    expect(result.value).toContain('not structured-cloneable')
+  })
+
+  it('exposes binding names that collide with Object.prototype as ordinary functions', async () => {
+    const { runtime } = await setup()
+    const result = await runtime.run({
+      program: 'return [await tools["__proto__"]({}), await tools["constructor"]({}), typeof tools["hasOwnProperty"]]',
+      // Computed keys: a literal `'__proto__': …` entry would SET the record's
+      // prototype instead of declaring a binding of that name.
+      bindings: tools({ ['__proto__']: async () => 'proto-ok', ['constructor']: async () => 'ctor-ok' }),
+    })
+    expect(result.value).toEqual(['proto-ok', 'ctor-ok', 'undefined'])
+  })
+})
+
+describe('WorkerCodeRuntime — seam misuse and lifecycle', () => {
+  it('rejects invalid binding globals loudly (identifier, reserved word, duplicate, console)', async () => {
+    const { runtime } = await setup()
+    const cases: [string, RegExp][] = [
+      ['not valid!', /not a usable identifier/],
+      ['await', /not a usable identifier/],
+      ['console', /duplicate binding global/],
+    ]
+    for (const [global, message] of cases) {
+      await expect(runtime.run({ program: 'return 1', bindings: [{ global, functions: {} }] })).rejects.toThrow(message)
+    }
+    await expect(runtime.run({
+      program: 'return 1',
+      bindings: [{ global: 'tools', functions: {} }, { global: 'tools', functions: {} }],
+    })).rejects.toThrow(/duplicate binding global/)
+  })
+
+  it('rejects config values that are not positive numbers', async () => {
+    const ctx = new Context()
+    await expect(ctx.plugin(WorkerCodeRuntime, { computeMs: -1 })).rejects.toThrow(/positive number/)
+  })
+
+  it('keeps runs isolated: no state survives from one run to the next', async () => {
+    const { runtime } = await setup()
+    await runtime.run({ program: 'globalThis.leak = "value"; return 1', bindings: [] })
+    const second = await runtime.run({ program: 'return typeof globalThis.leak', bindings: [] })
+    expect(second.value).toBe('undefined')
+  })
+
+  it('disposal aborts in-flight runs, awaits worker exit, and rejects later runs', async () => {
+    const ctx = new Context()
+    const fiber = await ctx.plugin(WorkerCodeRuntime)
+    const runtime = ctx.codeRuntime as WorkerCodeRuntime
+    const inflight: Promise<CodeRunResult> = runtime.run({ program: 'for (;;) {}', bindings: [] })
+    // Give the worker a moment to actually start spinning.
+    await new Promise(resolve => setTimeout(resolve, 200))
+    await fiber.dispose()
+    const result = await inflight
+    expect(result.error).toEqual({ kind: 'abort', message: 'runtime disposed' })
+    await expect(runtime.run({ program: 'return 1', bindings: [] })).rejects.toThrow(/after disposal/)
+  }, 15_000)
+
+  it('removes ctx.codeRuntime when the providing fiber disposes (HMR safety)', async () => {
+    const ctx = new Context()
+    const fiber = await ctx.plugin(WorkerCodeRuntime)
+    expect(ctx.get('codeRuntime')).toBeInstanceOf(WorkerCodeRuntime)
+    await fiber.dispose()
+    expect(ctx.get('codeRuntime')).toBeUndefined()
+  })
+})

packages/code-runtime/code-runtime-worker/tsconfig.json

@@ -0,0 +1,24 @@
+{
+  "extends": "../../../tsconfig.base.json",
+  "compilerOptions": {
+    "rootDir": "src",
+    "outDir": "lib/types"
+  },
+  "include": [
+    "src"
+  ],
+  "references": [
+    {
+      "path": "../../../vendor/cosmokit"
+    },
+    {
+      "path": "../../../vendor/cordis"
+    },
+    {
+      "path": "../../../vendor/schemastery"
+    },
+    {
+      "path": "../code-runtime"
+    }
+  ]
+}

packages/code-runtime/code-runtime-worker/tsdown.config.ts

@@ -0,0 +1,35 @@
+import { defineConfig } from 'tsdown'
+
+/**
+ * Package-shape override (see the root tsdown.config.ts): besides the
+ * default lib/index.js bundle, the worker BOOTSTRAP ships as its own
+ * sibling entry — `new Worker(new URL('./worker.js', import.meta.url))`
+ * loads it as a file, so it cannot be part of the index bundle. TWO
+ * single-entry builds, not one two-entry build: a multi-entry build emits
+ * the shared bootstrap module as a `lib/bootstrap-*.js` chunk both bundles
+ * import, which the package.json `files` whitelist (deliberately exact)
+ * would omit from the packed artifact — each single-entry build inlines its
+ * own bootstrap copy instead, keeping every shipped file self-contained.
+ */
+export default defineConfig([
+  {
+    entry: ['lib/types/index.js'],
+    outDir: 'lib',
+    format: ['esm'],
+    platform: 'node',
+    target: 'es2024',
+    fixedExtension: false,
+    dts: false,
+    clean: false,
+  },
+  {
+    entry: ['lib/types/worker.js'],
+    outDir: 'lib',
+    format: ['esm'],
+    platform: 'node',
+    target: 'es2024',
+    fixedExtension: false,
+    dts: false,
+    clean: false,
+  },
+])

packages/core/tools/README.md

@@ -1,6 +1,6 @@
 # dsh-tools
 
-Tool registry and execution pipeline. Tool plugins register their schemas and executors; the agent loop executes each call through `tools/pre-execute` (the allow/deny gate) → core dispatch → `tools/post-execute` (inspect/replace the result, attach context).
+Tool registry and execution pipeline. Tool plugins register their schemas and executors; the agent loop executes each call through `tools/pre-execute` (the allow/deny gate) → `tools/execute` (an around-dispatch wrapper for timeout/retry/metrics plugins) → `tools/post-execute` (inspect/replace the result, attach context).
 
 ## Service: `ToolRegistry` (ctx key: `tools`)
 
@@ -9,7 +9,7 @@ Tool registry and execution pipeline. Tool plugins register their schemas and ex
 - `ctx.tools.register(definition: ToolDefinition): () => void` Register a tool. Disposed with the calling fiber.
 - `ctx.tools.get(name: string): ToolDefinition | undefined`
 - `ctx.tools.schemas(): ToolSchema[]` Schemas of all registered tools (without the `execute` functions). The shipped tools' schemas are catalogued in [docs/tool-catalog.md](../../../docs/tool-catalog.md), generated by booting each tool plugin and harvesting this method (see [the tool-schema-catalog RFC](../../../docs/rfc/implemented/process/2026-07-02-tool-schema-catalog.md)).
-- `ctx.tools.execute(exec: ToolExecution): Promise<ToolExecutionResult>` Execute one tool call through the `tools/pre-execute` → dispatch → `tools/post-execute` pipeline.
+- `ctx.tools.execute(exec: ToolExecution): Promise<ToolExecutionResult>` Execute one tool call through the `tools/pre-execute` → `tools/execute` → `tools/post-execute` pipeline.
 
 ### Injected services
 
@@ -20,12 +20,13 @@ Tool registry and execution pipeline. Tool plugins register their schemas and ex
 | Event | Mode | Purpose |
 |---|---|---|
 | `tools/pre-execute` | waterfall | Allow/deny gate BEFORE a tool runs (sandbox, permission, hooks); returns `PreToolDecision` |
+| `tools/execute` | waterfall | Around-dispatch wrapper (timeout, retry, metrics): `(exec, next)` → the dispatched `ToolExecutionResult`; `next()` is dispatch-with-normalization |
 | `tools/post-execute` | waterfall | Inspect/replace the result AFTER a tool runs, attach context; returns `PostToolDecision` |
 | `tools/change` | emit | A tool was registered or unregistered |
 
 ### Key types
 
-- `ToolDefinition` — `ToolSchema` + `execute(args, exec): Promise<ContentBlock[] | { content: ContentBlock[]; meta? }>` (the bare array is the model-facing content; the object form additionally attaches an opaque, JSON-serializable `meta` presentation payload persisted on the `tool/result` event and handed back to `presentResult`), plus optional `presentCall(args)` / `presentResult(args, result)` for tool-owned UI presentation (see below).
+- `ToolDefinition` — `ToolSchema` + `execute(args, exec): Promise<ContentBlock[] | { content: ContentBlock[]; meta? }>` (the bare array is the model-facing content; the object form additionally attaches an opaque, JSON-serializable `meta` presentation payload persisted on the `tool/result` event and handed back to `presentResult`), plus optional `presentCall(args)` / `presentResult(args, result)` for tool-owned UI presentation (see below). It also carries an optional cooperative timeout budget `timeoutMs?: number` (ms) enforced by `@deepseek-ai/dsh-timeout-policy`, never sent to the model.
 - `ToolExecution` — one pending tool call: `{ callId, name, arguments, agent?, signal? }`.
 - `ToolExecutionResult` — outcome: `{ callId, content, isError, error?, additionalContext?, meta? }`. On failure with a `HarnessError`, `error: { name, code }` carries the structured failure class alongside the model-facing text (the loop forwards it onto the `tool/result` session event for retry/sandbox plugins and replay). `additionalContext` (a `HookContext`) ferries any `tools/post-execute` context up to the loop, which buffers it and appends it as a `context/message` after all `tool/result`s in the step. `meta` is the tool's opaque presentation payload from a successful `execute` (the object return form); the loop forwards it onto the `tool/result` session event for result-card rendering.
 - `PreToolDecision` — `{kind:'allow'}` | `{kind:'deny', reason}` | `{kind:'ask', reason?}`. Input rewrite (changing `arguments`) is deliberately NOT offered (it would desync the pre-execution audit/history/UI from what ran — its own proposed RFC); `ask` degrades to `deny` until the permission system lands.
@@ -35,7 +36,7 @@ Tool registry and execution pipeline. Tool plugins register their schemas and ex
 ### Extension points
 
 - Tool plugins call `ctx.tools.register()` — schemas flow into the assembly automatically.
-- `tools/pre-execute` is the allow/deny gate (sandbox, permission, hooks): listeners receive `(exec, next)` and call `next()` to delegate to the default (allow) or return a `PreToolDecision` to short-circuit; a `deny`/`ask` skips dispatch and yields an `isError` result. `tools/post-execute` is the inspect/transform seam: `(exec, result, next)` → a `PostToolDecision` that can replace content, block with feedback, or attach `additionalContext`. Core dispatch sits between them as plain code; the tool body keeps its own try/catch so a thrown tool still reaches `post-execute` as an `isError`. Both follow the typed-Decision idiom shared with the `agent/*` interception seams (see [`dsh-agent`](../agent/README.md)).
+- `tools/pre-execute` is the allow/deny gate (sandbox, permission, hooks): listeners receive `(exec, next)` and call `next()` to delegate to the default (allow) or return a `PreToolDecision` to short-circuit; a `deny`/`ask` skips dispatch and yields an `isError` result. `tools/execute` is the around-dispatch seam (timeout, retry, metrics): listeners receive `(exec, next)` and call `next()` to delegate to core dispatch (returning its `ToolExecutionResult`, optionally wrapped), or return a replacement result to short-circuit dispatch; the base `next()` IS dispatch-with-normalization, so `await next()` already yields an `isError` result for a thrown/unknown tool (never a raw throw). A wrapper mutates `exec` in place before `next()` — e.g. replacing `exec.signal` with a per-call deadline — because cordis `next()` ignores passed arguments. `tools/post-execute` is the inspect/transform seam: `(exec, result, next)` → a `PostToolDecision` that can replace content, block with feedback, or attach `additionalContext`. Core dispatch is the base of the `tools/execute` waterfall; the tool body keeps its own try/catch so a thrown tool still reaches `post-execute` as an `isError`. All follow the typed-Decision idiom shared with the `agent/*` interception seams (see [`dsh-agent`](../agent/README.md)); `@deepseek-ai/dsh-timeout-policy` is the reference `tools/execute` wrapper.
 - MCP servers: one plugin per server, discover tools, call `ctx.tools.register()` with the server's schemas.
 
 ### Typed tool parameter schemas
@@ -71,6 +72,8 @@ A `defineTool` tool also **validates the model-generated arguments against its `
 
 See `defineTool`, `validateArgs`, `ToolArgsError`, `SchemaSpec`, `InferArgs`, and `schemaSpecToJsonSchema` in the public API for details.
 
+`defineTool` also validates an optional `timeoutMs` at definition time when present: it must be a positive finite number, or the helper throws — the budget is attached to the produced `ToolDefinition` (for `@deepseek-ai/dsh-timeout-policy`) and never reaches the model.
+
 ### Structured-output schema subset
 
 A separate vocabulary for callers that DEMAND a machine-readable value from an agent — the subagent seam's `SubagentStartRequest.outputSchema` (and, by extension, a workflow's `agent({ schema })`). Unlike `SchemaSpec` (the author-facing DSL for tool parameters), a `StructuredOutputSchema` is an object-rooted **raw JSON Schema subset** as data: it travels verbatim to the model as a forced tool's `parameters`, and the produced value is validated against it.

packages/core/tools/src/index.ts

@@ -1,9 +1,10 @@
 /**
  * Tool registry and execution pipeline. Plugins register tools; the registry
  * feeds schemas into the system prompt, and `execute()` dispatches each call
- * through `tools/pre-execute` (the allow/deny gate) → core dispatch →
- * `tools/post-execute` (inspect/replace the result, attach context) for
- * sandbox, permission, and hook plugins to gate or transform a call.
+ * through `tools/pre-execute` (the allow/deny gate) → `tools/execute` (an
+ * around-dispatch wrapper for timeout/retry/metrics plugins) → `tools/post-execute`
+ * (inspect/replace the result, attach context) for sandbox, permission, and hook
+ * plugins to gate or transform a call.
  *
  * @module @deepseek-ai/dsh-tools
  */
@@ -74,17 +75,37 @@ declare module 'cordis' {
      * @mode waterfall
      */
     'tools/pre-execute'(this: ToolRegistry, exec: ToolExecution, next: () => Promise<PreToolDecision>): Promise<PreToolDecision>
+    /**
+     * Around-dispatch waterfall wrapping the registry's core tool dispatch,
+     * between the `tools/pre-execute` gate and the `tools/post-execute` seam. A
+     * listener receives `(exec, next)`: call `next()` to delegate to dispatch
+     * (returning its {@link ToolExecutionResult}, optionally wrapped), or return a
+     * replacement result without calling `next()` to short-circuit dispatch. The
+     * base `next()` IS the dispatch-with-normalization thunk — a thrown tool (or
+     * unknown tool) is already normalized to an `isError` result by the time a
+     * listener's `await next()` returns, so a wrapper never sees a raw throw from
+     * the tool body. This is the seam a timeout/retry/metrics plugin wraps: it can
+     * mutate `exec` (e.g. replace `exec.signal` with a per-call deadline) BEFORE
+     * `next()` and inspect the result AFTER. (Cordis `next()` ignores any passed
+     * arguments and re-invokes downstream with the shared payload, so a wrapper
+     * mutates `exec` in place rather than passing a new object to `next()`.)
+     * Multiple listeners compose by registration order — an outer one wraps the
+     * inner ones plus dispatch.
+     * @param exec - the allowed call about to dispatch (name, parsed arguments, caller agent, signal).
+     * @mode waterfall
+     */
+    'tools/execute'(this: ToolRegistry, exec: ToolExecution, next: () => Promise<ToolExecutionResult>): Promise<ToolExecutionResult>
     /**
      * Waterfall AFTER a tool runs — where hook plugins inspect the result and
      * accept it (optionally REPLACING the model-facing content, and/or attaching
      * `additionalContext` for the next request) or block it with corrective
      * `feedback` (Claude Code's `PostToolUse`). Listeners receive
      * `(exec, result, next)`: call `next()` to delegate to the default (accept
-     * unchanged), or return a {@link PostToolDecision} to override. The core tool
-     * dispatch sits between the two waterfalls as plain code, all inside
-     * `execute`'s outer try/catch (and the tool body keeps its own inner
-     * try/catch, so a thrown tool still reaches `post-execute` as an `isError`
-     * result).
+     * unchanged), or return a {@link PostToolDecision} to override. Core tool
+     * dispatch runs earlier as the base `next()` of the `tools/execute`
+     * waterfall, all inside `execute`'s outer try/catch (and the tool body keeps
+     * its own inner try/catch, so a thrown tool still reaches `post-execute` as an
+     * `isError` result).
      * @param exec - the call that just ran (name, parsed arguments, caller agent).
      * @param result - the dispatch outcome a listener may accept, replace, or block.
      * @mode waterfall
@@ -117,6 +138,14 @@ export type ToolExecuteReturn = ContentBlock[] | { content: ContentBlock[]; meta
 /** A registered tool: its schema plus the execution function. */
 export interface ToolDefinition extends ToolSchema {
   execute(args: unknown, exec: ToolExecution): Promise<ToolExecuteReturn>
+  /**
+   * Cooperative tool-call timeout budget in milliseconds. Omit for no deadline.
+   * Enforced by `@deepseek-ai/dsh-timeout-policy` (a `tools/execute` wrapper); it
+   * is NEVER sent to the model — `schemas()` whitelists only name/description/
+   * parameters. Declaring it asserts this tool forwards `exec.signal` to a
+   * cooperative implementation that can reach quiescence when the signal aborts.
+   */
+  timeoutMs?: number
   /**
    * Optional: how to present the PENDING state of one call in a UI, derived from
    * the call's `args` (parsed arguments, `unknown` — the tool validates/narrows
@@ -271,7 +300,7 @@ function errorInfo(error: unknown): ToolErrorInfo | undefined {
 
 /**
  * Tool registry (`ctx.tools`): tool plugins register definitions; the agent
- * loop executes calls through the `tools/pre-execute` → dispatch →
+ * loop executes calls through the `tools/pre-execute` → `tools/execute` →
  * `tools/post-execute` pipeline. The registry contributes its schemas into the
  * system-prompt assembly.
  */
@@ -345,18 +374,20 @@ export class ToolRegistry extends Service {
   }
 
   /**
-   * Execute one tool call through the `tools/pre-execute` → dispatch →
-   * `tools/post-execute` pipeline. The two waterfalls are the gate (allow/deny)
-   * and the inspect/transform seam; core dispatch sits between them as plain
-   * code. The whole thing is wrapped in one outer try/catch so a throwing
-   * listener (in either waterfall) becomes an `isError` result instead of
-   * failing the turn; the tool body ALSO keeps its own inner try/catch, so a
-   * thrown tool becomes an `isError` result that `post-execute` listeners can
-   * still inspect. If the tool is not registered, the result is an `isError`
-   * carrying a `UNKNOWN_TOOL` structured error. A thrown {@link HarnessError}
-   * surfaces its `{ name, code }` on the result.
+   * Execute one tool call through the `tools/pre-execute` → `tools/execute`
+   * (around dispatch) → `tools/post-execute` pipeline. `pre-execute` is the gate
+   * (allow/deny), `tools/execute` wraps core dispatch (a timeout/retry/metrics
+   * seam), and `post-execute` is the inspect/transform seam; core dispatch sits
+   * as the base `next()` of the `tools/execute` waterfall. The whole thing is
+   * wrapped in one outer try/catch so a throwing listener (in any waterfall)
+   * becomes an `isError` result instead of failing the turn; the tool body ALSO
+   * keeps its own inner try/catch, so a thrown tool becomes an `isError` result
+   * that `tools/execute` and `post-execute` listeners can still inspect. If the
+   * tool is not registered, the result is an `isError` carrying a `UNKNOWN_TOOL`
+   * structured error. A thrown {@link HarnessError} surfaces its `{ name, code }`
+   * on the result.
    * @param exec - the call to run (name, parsed arguments, caller agent, signal).
-   * @returns the final result after both waterfalls; failures resolve as
+   * @returns the final result after every waterfall; failures resolve as
    *   `isError` results, never rejections.
    */
   async execute(exec: ToolExecution): Promise<ToolExecutionResult> {
@@ -382,23 +413,30 @@ export class ToolRegistry extends Service {
         return await this.postExecute(exec, denied)
       }
 
-      // --- Core dispatch (plain code between the waterfalls). The tool body's
-      // own try/catch turns a throw into an isError result so post-execute can
-      // inspect it; an unknown tool routes through the same catch. ---
-      let result: ToolExecutionResult
-      try {
-        const tool = this.store.get(exec.name)
-        if (!tool) throw new ToolNotFoundError(exec.name)
-        // Normalize the two `execute` return shapes: a bare ContentBlock[] (no
-        // meta) or a { content, meta } object (a tool attaching a private
-        // presentation payload). An array IS the content; the object carries it.
-        const returned = await tool.execute(exec.arguments, exec)
-        const content = Array.isArray(returned) ? returned : returned.content
-        const meta = Array.isArray(returned) ? undefined : returned.meta
-        result = { callId: exec.callId, content, isError: false, ...meta !== undefined ? { meta } : {} }
-      } catch (error: unknown) {
-        result = toolErrorResult(exec.callId, error)
-      }
+      // --- Around-dispatch: tools/execute. The base `next` is the dispatch-
+      // with-normalization thunk — the tool body's own try/catch turns a throw
+      // into an isError result so a wrapper (and post-execute) can inspect it;
+      // an unknown tool routes through the same catch. A `tools/execute` listener
+      // (e.g. a timeout plugin) wraps this thunk: it may mutate `exec` before
+      // delegating and inspect the normalized result after. ---
+      const result = await this.ctx.waterfall(
+        this, 'tools/execute', exec,
+        async (): Promise<ToolExecutionResult> => {
+          try {
+            const tool = this.store.get(exec.name)
+            if (!tool) throw new ToolNotFoundError(exec.name)
+            // Normalize the two `execute` return shapes: a bare ContentBlock[] (no
+            // meta) or a { content, meta } object (a tool attaching a private
+            // presentation payload). An array IS the content; the object carries it.
+            const returned = await tool.execute(exec.arguments, exec)
+            const content = Array.isArray(returned) ? returned : returned.content
+            const meta = Array.isArray(returned) ? undefined : returned.meta
+            return { callId: exec.callId, content, isError: false, ...meta !== undefined ? { meta } : {} }
+          } catch (error: unknown) {
+            return toolErrorResult(exec.callId, error)
+          }
+        },
+      )
 
       return await this.postExecute(exec, result)
     } catch (error: unknown) {

packages/core/tools/src/schema.ts

@@ -295,6 +295,13 @@ export interface DefineToolOptions<S extends SchemaSpec> {
    * standard JSON Schema at runtime.
    */
   parameters: S
+  /**
+   * Optional cooperative tool-call timeout budget in milliseconds. When given it
+   * must be a positive finite number; it is attached to the produced
+   * {@link ToolDefinition} for `@deepseek-ai/dsh-timeout-policy` to enforce and
+   * is never sent to the model.
+   */
+  timeoutMs?: number
   /**
    * Tool execution function. `args` is typed as {@link InferArgs<S>} — zero
    * casts needed. Returns either a bare {@link ContentBlock}`[]` (model-facing
@@ -362,10 +369,14 @@ export function defineTool<S extends SchemaSpec>(options: DefineToolOptions<S>):
   const userPresentCall = options.presentCall
   // eslint-disable-next-line @typescript-eslint/unbound-method
   const userPresentResult = options.presentResult
+  if (options.timeoutMs !== undefined && (!Number.isFinite(options.timeoutMs) || options.timeoutMs <= 0)) {
+    throw new Error(`defineTool(${options.name}): timeoutMs must be a positive finite number`)
+  }
   const tool: ToolDefinition = {
     name: options.name,
     description: options.description,
     parameters: schemaSpecToJsonSchema(options.parameters) as unknown as Record<string, unknown>,
+    ...(options.timeoutMs !== undefined ? { timeoutMs: options.timeoutMs } : {}),
     async execute(args: unknown, exec: ToolExecution): Promise<ToolExecuteReturn> {
       // Validate the model-generated args before the typed body runs. On
       // mismatch we throw ToolArgsError; the registry turns it into an

packages/core/tools/tests/gen-tool-catalog.spec.ts

@@ -35,7 +35,7 @@ describe('gen-tool-catalog collectToolCatalog', () => {
   it('boots every shipped tool package and harvests its model-facing schemas', async () => {
     const catalog = await collectToolCatalog()
     const names = catalog.flatMap(entry => entry.schemas.map(s => s.name)).sort()
-    expect(names).toEqual(['bash', 'edit', 'read', 'subagent', 'task_kill', 'task_list', 'task_output', 'todo_write', 'web_fetch', 'web_search', 'write'])
+    expect(names).toEqual(['ask_user_question', 'bash', 'edit', 'read', 'subagent', 'task_kill', 'task_list', 'task_output', 'todo_write', 'web_fetch', 'web_search', 'write'])
     // Every tool carries a JSON-Schema `parameters` object (what the model sees).
     for (const entry of catalog) {
       for (const schema of entry.schemas) {

packages/core/tools/tests/tools.spec.ts

@@ -5,6 +5,7 @@ import SystemPrompt from '@deepseek-ai/dsh-system-prompt'
 import ToolRegistry, {
   defineTool, schemaSpecToJsonSchema, validateArgs, ToolArgsError, ToolNotFoundError,
   type InferArgs, type SchemaSpec, type PreToolDecision, type PostToolDecision,
+  type ToolExecution, type ToolExecutionResult,
 } from '@deepseek-ai/dsh-tools'
 
 async function setup() {
@@ -62,6 +63,17 @@ describe('ToolRegistry', () => {
     expect(schema.execute).toBeUndefined()
   })
 
+  it('schemas() excludes timeoutMs — the budget must never reach the model', async () => {
+    const ctx = await setup()
+    ctx.tools.register(defineTool({
+      name: 'budgeted', description: 'has a budget', parameters: {}, timeoutMs: 5_000,
+      async execute() { return [{ type: 'text' as const, text: 'ok' }] },
+    }))
+    const schema = ctx.tools.schemas().find(s => s.name === 'budgeted')
+    expect(schema).toBeDefined()
+    expect('timeoutMs' in (schema as object)).toBe(false)
+  })
+
   it('executes a tool and returns its content', async () => {
     const ctx = await setup()
     ctx.tools.register(echoTool)
@@ -272,6 +284,151 @@ describe('ToolRegistry', () => {
     expect(order).toEqual(['pre:before', 'pre:after', 'post:before', 'post:after'])
   })
 
+  it('runs tools/execute after an allowed pre-execute, around dispatch, and before post-execute', async () => {
+    const ctx = await setup()
+    const order: string[] = []
+    ctx.tools.register(defineTool({
+      name: 'traced',
+      description: 'echo',
+      parameters: { text: { type: 'string' } },
+      async execute(args) {
+        order.push('dispatch')
+        return [{ type: 'text' as const, text: args.text ?? '' }]
+      },
+    }))
+
+    ctx.on('tools/pre-execute', async (_exec, next) => { order.push('pre'); return next() })
+    ctx.on('tools/execute', async (_exec: ToolExecution, next: () => Promise<ToolExecutionResult>): Promise<ToolExecutionResult> => {
+      order.push('execute:before')
+      const result = await next()
+      order.push('execute:after')
+      return result
+    })
+    ctx.on('tools/post-execute', async (_exec, _result, next) => { order.push('post'); return next() })
+
+    const result = await ctx.tools.execute({ callId: CallId('c1'), name: 'traced', arguments: { text: 'hi' } })
+    expect(result).toEqual({ callId: CallId('c1'), content: [{ type: 'text', text: 'hi' }], isError: false })
+    // The around seam wraps dispatch; pre gates before it, post runs over its result.
+    expect(order).toEqual(['pre', 'execute:before', 'dispatch', 'execute:after', 'post'])
+  })
+
+  it('a pre-execute deny short-circuits before tools/execute (the seam never runs)', async () => {
+    const ctx = await setup()
+    ctx.tools.register(echoTool)
+
+    let entered = false
+    ctx.on('tools/pre-execute', async (_exec, _next): Promise<PreToolDecision> => ({ kind: 'deny', reason: 'nope' }))
+    ctx.on('tools/execute', async (_exec: ToolExecution, next: () => Promise<ToolExecutionResult>): Promise<ToolExecutionResult> => {
+      entered = true
+      return next()
+    })
+
+    const result = await ctx.tools.execute({ callId: CallId('c1'), name: 'echo', arguments: { text: 'hi' } })
+    expect(result.isError).toBe(true)
+    expect(result.content[0]).toMatchObject({ text: 'Error: nope' })
+    expect(entered).toBe(false) // a denied call never enters the around-dispatch seam
+  })
+
+  it('a thrown tool is normalized to an isError result BEFORE a tools/execute listener sees next()', async () => {
+    const ctx = await setup()
+    ctx.tools.register({
+      ...echoTool,
+      name: 'boom',
+      async execute() { throw new HarnessError('kaboom', 'BOOM') },
+    })
+
+    let seen: { isError: boolean; error?: unknown } | undefined
+    ctx.on('tools/execute', async (_exec: ToolExecution, next: () => Promise<ToolExecutionResult>): Promise<ToolExecutionResult> => {
+      const result = await next()
+      // The base next() IS dispatch-with-normalization: the wrapper sees the
+      // normalized isError result, never a raw throw from the tool body.
+      seen = { isError: result.isError, error: result.error }
+      return result
+    })
+
+    const result = await ctx.tools.execute({ callId: CallId('c1'), name: 'boom', arguments: {} })
+    expect(seen).toEqual({ isError: true, error: { name: 'HarnessError', code: 'BOOM' } })
+    expect(result.isError).toBe(true)
+    expect(result.content[0]).toMatchObject({ text: 'Error: kaboom' })
+  })
+
+  it('a thrown tool normalized inside tools/execute still reaches post-execute', async () => {
+    const ctx = await setup()
+    ctx.tools.register({
+      ...echoTool,
+      name: 'boom',
+      async execute() { throw new Error('exploded') },
+    })
+
+    let postSaw: boolean | undefined
+    ctx.on('tools/execute', async (_exec: ToolExecution, next: () => Promise<ToolExecutionResult>): Promise<ToolExecutionResult> => next())
+    ctx.on('tools/post-execute', async (_exec, result, next) => {
+      postSaw = result.isError
+      return next()
+    })
+
+    const result = await ctx.tools.execute({ callId: CallId('c1'), name: 'boom', arguments: {} })
+    expect(postSaw).toBe(true) // the normalized isError still flows through post-execute
+    expect(result.isError).toBe(true)
+    expect(result.content[0]).toMatchObject({ text: 'Error: exploded' })
+  })
+
+  it('a tools/execute listener can replace exec.signal for the dispatched tool (deadline pattern)', async () => {
+    const ctx = await setup()
+    let seenSignal: AbortSignal | undefined
+    ctx.tools.register({
+      ...echoTool,
+      name: 'signal-probe',
+      async execute(_args, exec) {
+        seenSignal = exec.signal
+        return [{ type: 'text' as const, text: 'ok' }]
+      },
+    })
+
+    const upstream = new AbortController().signal
+    const replacement = new AbortController().signal
+    ctx.on('tools/execute', async (exec: ToolExecution, next: () => Promise<ToolExecutionResult>): Promise<ToolExecutionResult> => {
+      expect(exec.signal).toBe(upstream)
+      // Cordis next() ignores passed arguments, so a wrapper mutates exec in
+      // place (the documented "mutate the shared object, then delegate" idiom).
+      exec.signal = replacement
+      return next()
+    })
+
+    await ctx.tools.execute({ callId: CallId('c1'), name: 'signal-probe', arguments: {}, signal: upstream })
+    expect(seenSignal).toBe(replacement) // dispatch saw the wrapper's replacement, not the upstream
+  })
+
+  it('a tools/execute listener can short-circuit dispatch by returning a result without next()', async () => {
+    const ctx = await setup()
+    let dispatched = false
+    ctx.tools.register({
+      ...echoTool,
+      name: 'never-runs',
+      async execute() { dispatched = true; return [] },
+    })
+
+    ctx.on('tools/execute', async (exec: ToolExecution, _next: () => Promise<ToolExecutionResult>): Promise<ToolExecutionResult> =>
+      ({ callId: exec.callId, content: [{ type: 'text', text: 'short-circuited' }], isError: false }))
+
+    const result = await ctx.tools.execute({ callId: CallId('c1'), name: 'never-runs', arguments: {} })
+    expect(dispatched).toBe(false) // returning without next() skips core dispatch
+    expect(result.content[0]).toMatchObject({ text: 'short-circuited' })
+  })
+
+  it('returns an isError result when a tools/execute listener throws', async () => {
+    const ctx = await setup()
+    ctx.tools.register(echoTool)
+    ctx.on('tools/execute', async () => { throw new Error('wrapper broke') })
+
+    const result = await ctx.tools.execute({ callId: CallId('c1'), name: 'echo', arguments: { text: 'hi' } })
+    expect(result).toEqual({
+      callId: CallId('c1'),
+      content: [{ type: 'text', text: 'Error: wrapper broke' }],
+      isError: true,
+    })
+  })
+
   it('returns an isError result when a tools/pre-execute listener throws', async () => {
     const ctx = await setup()
     ctx.tools.register(echoTool)
@@ -989,6 +1146,38 @@ describe('defineTool validation (the runtime-validation RFC, part 1)', () => {
     const result = await ctx.tools.execute({ callId: CallId('c1'), name: 'raw', arguments: {} })
     expect(result.isError).toBe(false)
   })
+
+  it('attaches a positive-finite timeoutMs to the definition', () => {
+    const tool = defineTool({
+      name: 'x', description: 'd', parameters: {}, timeoutMs: 30_000,
+      async execute() { return [{ type: 'text' as const, text: 'ok' }] },
+    })
+    expect(tool.timeoutMs).toBe(30_000)
+  })
+
+  it('omits timeoutMs when not declared', () => {
+    const tool = defineTool({
+      name: 'x', description: 'd', parameters: {},
+      async execute() { return [{ type: 'text' as const, text: 'ok' }] },
+    })
+    expect(tool.timeoutMs).toBeUndefined()
+  })
+
+  it('throws when timeoutMs is zero or negative', () => {
+    const make = (ms: number) => defineTool({
+      name: 'x', description: 'd', parameters: {}, timeoutMs: ms,
+      async execute() { return [{ type: 'text' as const, text: 'ok' }] },
+    })
+    expect(() => make(0)).toThrow('timeoutMs must be a positive finite number')
+    expect(() => make(-5)).toThrow('positive finite number')
+  })
+
+  it('throws when timeoutMs is non-finite', () => {
+    expect(() => defineTool({
+      name: 'x', description: 'd', parameters: {}, timeoutMs: Infinity,
+      async execute() { return [{ type: 'text' as const, text: 'ok' }] },
+    })).toThrow('positive finite number')
+  })
 })
 
 describe('defineTool presentation (presentCall / presentResult)', () => {

packages/fs/README.md

@@ -10,3 +10,7 @@ The filesystem stack: a provider seam (text IO + atomic mutation with an optiona
 | `tool-fs/` | Model-facing `read`/`write`/`edit` tools AND the executor (reads via `ctx.fs`, owns read windowing, dispatches `fs/*`) | (registers on `ctx.tools`) |
 
 The interface lives at `fs/fs/`. A sandboxed, remote, or project-scoped filesystem backend can replace `fs-local` without touching the seam, the policy gate, or the model-facing tool schemas. The policy (`fs-policy/`) is a plugin that participates only through the `fs/*` event gate, not a service the tool injects — so dropping it gracefully loses the policy and leaves the unconstrained bare provider rather than breaking the tool. A deployment that loads `tool-fs/` is expected to also load it.
+
+## No timeouts on file IO
+
+`read`/`write`/`edit` take **no** `timeoutMs`, and the provider seam arms no deadline — unlike bash and web, which consume [`@deepseek-ai/dsh-timeout`](../util/timeout/README.md). A local syscall is best-effort-abortable at most: a timeout could not force an in-progress `fsync`/`rename` to stop, so a deadline here would be a knob that cannot deliver on its promise. Adding one would also be an implicit default in the exact place explicit-over-implicit forbids. Both reference agents (Claude Code, Codex) leave file IO untimed for the same reason; cancellation still propagates through the tool-execution signal for best-effort abort at syscall boundaries.

packages/guard/README.md

@@ -0,0 +1,9 @@
+# guard/ — loop-hygiene guard family
+
+Behavioral guard plugins that watch the agent loop for unproductive patterns and nudge the model back on course. A single **product** package — there is no interface/implementation seam here, because a guard is a self-contained consumer of existing core seams (`tools/post-execute`, `agent/prompt-submit`, `agent/status`), not a swappable capability.
+
+| Package | Role | ctx key |
+|---|---|---|
+| `repeat-tool-guard/` | Advisory reminders when an agent loops on identical tool calls | (listens on `ctx.tools`' waterfalls) |
+
+Reminders travel as `additionalContext` on the `tools/post-execute` decision; the agent loop appends them as logged `context/message` events after the step's tool results (see [the tools package](../core/tools)), so everything a guard says to the model is reconstructable from the session log.

packages/guard/repeat-tool-guard/README.md

@@ -0,0 +1,37 @@
+# @deepseek-ai/dsh-repeat-tool-guard
+
+An advisory loop-breaker, not a model-facing tool: it never appears in the tool list, never vetoes or rewrites a call, and adds exactly one behavior — it watches each agent's stream of tool calls, counts runs of consecutive calls to the same tool with identical canonicalized arguments, and at configured run lengths injects an escalating advisory reminder telling the model to stop repeating itself, re-read the last result, and either change approach or conclude. The decision (retry differently, gather more evidence, or finish) stays entirely with the model: a legitimately repeated call is delayed by nothing and blocked by nothing. Decision record: [the repeat-tool-guard RFC](../../../docs/rfc/implemented/feature/2026-07-08-repeat-tool-guard.md).
+
+## Config
+
+```yaml
+- id: repeat-tool-guard
+  name: '@deepseek-ai/dsh-repeat-tool-guard'
+  config:
+    thresholds: [3, 5, 8]        # default; consecutive counts that trigger a reminder
+    include: []                  # tool-name patterns to track; empty ⇒ all tools
+    exclude: [todo_write]        # tool-name patterns transparent to the chain
+    argumentsPreviewChars: 500   # default; cap on arguments quoted in the detailed reminder
+```
+
+`thresholds` fails loud at plugin load: an empty list, a non-integer, a value below 2, or a duplicate throws, never a silent fall-back to defaults; `argumentsPreviewChars` equally rejects anything but an integer >= 1. The list is normalized to ascending order; the FIRST threshold delivers a short generic nudge, every later threshold delivers the detailed form naming the tool, the run length, and the canonical arguments — head-truncated at `argumentsPreviewChars` with an omitted-count marker, so a looping `write`/`edit` payload cannot ride into the next request unbounded (the chain key always compares the FULL canonical string; the cap bounds the reminder, never the detection).
+
+`include`/`exclude` entries support `*` wildcards and are predicates over whatever tools exist at call time, not references to registry entries — a pattern matching no currently registered tool is NOT an error (`exclude: [mcp_*]` stays valid in a deployment that loads no MCP tools), unlike `toolOrder`'s referent check.
+
+## Chain semantics
+
+The chain key is `(tool name, canonical arguments)` — canonicalization is a deep key-sort plus `JSON.stringify`, so argument objects differing only in property order count as identical. A call identical to the previous tracked call increments the agent's consecutive counter; a different tracked call resets it to 1.
+
+- **Untracked calls are transparent to the chain.** A call excluded by `include`/`exclude` neither increments nor resets the counter, so `grep X → todo_write → grep X` still counts as two consecutive `grep X` when `todo_write` is excluded. This is what makes exclusion useful: bookkeeping tools interleaved into a loop must not launder it.
+- **Denied calls count.** Detection sits on `tools/post-execute`, which also runs for calls a `tools/pre-execute` listener denied — a model hammering a denied call is exactly the loop worth breaking.
+- **Calls without an agent are ignored.** A direct `ctx.tools.execute()` caller has no model to remind and no `AgentId` to key on.
+- **Per-agent keying.** The tool registry is context-level and subagents interleave through the same waterfall, so chains are keyed by `AgentId`; one agent's repetition never trips another's reminder. A user prompt (`agent/prompt-submit`) resets the submitting agent's chain; agent disposal drops its state.
+- **In-memory only.** A session resumed from persistence starts with a fresh chain — the guard is a heuristic nudge, not a logged invariant, later reminders are the accepted cost.
+
+## Reminder delivery
+
+Reminders ride the post-execute decision's `additionalContext` (source `{kind: 'plugin', plugin: 'repeat-tool-guard'}`), never a `content` replacement: the `tool/result` event stays the tool's own output for audit. The loop buffers the context and appends it as a `context/message` after the step's tool results, which the session renders as the tagged synthetic-user envelope — so the reminder is model-visible, source-attributed, and reconstructable from the session log with no new session event. The guard always delegates via `next()` and folds its reminder onto the downstream decision (both variants — a blocked call still gets the nudge); when a downstream listener attached its own `additionalContext`, the fold concatenates content and carries the guard's `source` (a `HookContext` holds one `MessageSource`; `source.kind` is what framing depends on).
+
+## Testing
+
+Unit suites drive a real agent loop against a mock adapter (no network) and cover the chain semantics above to per-file 100%. The snapshot tier owns the transcript surface: a scripted-replay scenario repeats a call five times and pins both reminder tiers (gentle at 3, detailed at 5) as `context/message`s in the ACP transcript.

packages/guard/repeat-tool-guard/package.json

@@ -0,0 +1,41 @@
+{
+  "name": "@deepseek-ai/dsh-repeat-tool-guard",
+  "description": "Repeat-tool-call guard plugin: advisory reminders when an agent loops on identical tool calls",
+  "version": "0.0.1",
+  "private": true,
+  "type": "module",
+  "main": "lib/index.js",
+  "types": "lib/types/index.d.ts",
+  "exports": {
+    ".": {
+      "types": "./lib/types/index.d.ts",
+      "default": "./lib/index.js"
+    },
+    "./src/*": "./src/*",
+    "./package.json": "./package.json"
+  },
+  "files": [
+    "lib/index.js",
+    "lib/types/**/*.d.ts",
+    "lib/types/**/*.d.ts.map",
+    "src"
+  ],
+  "license": "BSD-3-Clause",
+  "dependencies": {
+    "schemastery": "^3.18.0"
+  },
+  "peerDependencies": {
+    "@deepseek-ai/dsh-agent": "^0.0.1",
+    "@deepseek-ai/dsh-tools": "^0.0.1",
+    "cordis": "^4.0.0-rc.6"
+  },
+  "devDependencies": {
+    "@deepseek-ai/dsh-agent": "workspace:^",
+    "@deepseek-ai/dsh-agent-loop": "workspace:^",
+    "@deepseek-ai/dsh-llm": "workspace:^",
+    "@deepseek-ai/dsh-session": "workspace:^",
+    "@deepseek-ai/dsh-system-prompt": "workspace:^",
+    "@deepseek-ai/dsh-tools": "workspace:^",
+    "cordis": "^4.0.0-rc.6"
+  }
+}

packages/guard/repeat-tool-guard/src/index.ts

@@ -0,0 +1,268 @@
+/**
+ * Repeat-tool-call guard: advisory loop-breaker for agents stuck re-issuing
+ * the same tool call with identical arguments.
+ *
+ * Not a model-facing tool — it registers no tool, never vetoes or rewrites a
+ * call, and adds exactly one behavior: watch each agent's stream of tool calls
+ * through the `tools/post-execute` waterfall, count runs of consecutive calls
+ * to the same tool with identical canonicalized arguments, and at configured
+ * run lengths fold an escalating advisory reminder onto the decision's
+ * `additionalContext`. The loop appends that context as a logged
+ * `context/message` after the step's tool results, so the reminder is
+ * model-visible, source-attributed, and reconstructable from the session log
+ * with no new session event. Decision record:
+ * docs/rfc/implemented/feature/2026-07-08-repeat-tool-guard.md.
+ *
+ * ```yaml
+ * - id: repeat-tool-guard
+ *   name: '@deepseek-ai/dsh-repeat-tool-guard'
+ *   config:
+ *     thresholds: [3, 5, 8]   # consecutive counts that trigger a reminder
+ *     include: []             # tool-name patterns to track; empty = all tools
+ *     exclude: [todo_write]   # tool-name patterns transparent to the chain
+ * ```
+ *
+ * Chain state is keyed per {@link AgentId} — the tool registry is a
+ * context-level singleton whose waterfalls interleave every agent's calls, so
+ * a shared counter would let one agent's repetition trip another's reminder.
+ * State is in-memory only: a session resumed from persistence starts with a
+ * fresh chain (the guard is a heuristic nudge, not a logged invariant).
+ *
+ * Plugin export shape: named exports, NO default. The cordis Loader's
+ * `unwrapExports` does `exports.default ?? exports`, so a stray default would
+ * collapse the module to the bare `apply` (see docs/postmortem/0001).
+ *
+ * @module @deepseek-ai/dsh-repeat-tool-guard
+ */
+
+import type { Context } from 'cordis'
+import z from 'schemastery'
+import type { AgentId, HookContext, PromptDecision } from '@deepseek-ai/dsh-agent'
+import type { MessageSource } from '@deepseek-ai/dsh-llm'
+import type { PostToolDecision, ToolExecution } from '@deepseek-ai/dsh-tools'
+
+export const name = 'repeat-tool-guard'
+
+/**
+ * Plugin config, validated by the same-named schemastery schema plus the
+ * load-time checks in `apply` (misconfiguration fails loud: an empty
+ * `thresholds` list, a non-integer, a value below 2, or a duplicate throws at
+ * plugin load, never a silent fall-back). `include`/`exclude` entries are
+ * `*`-wildcard predicates over tool names at call time, not references to
+ * registry entries — a pattern matching no currently registered tool is valid
+ * (`exclude: [mcp_*]` must stay legal in a deployment that loads no MCP tools).
+ */
+export interface Config {
+  /** Consecutive-repeat counts that trigger a reminder (default `[3, 5, 8]`). */
+  thresholds?: number[]
+  /** Tool-name patterns to track; empty means every tool is tracked. */
+  include?: string[]
+  /** Tool-name patterns transparent to the chain (neither count nor reset). */
+  exclude?: string[]
+  /**
+   * Maximum characters of canonical arguments quoted in the DETAILED reminder
+   * (default 500). Large payloads (a `write` body, a long command) would
+   * otherwise ride into the next request unbounded — precisely in a loop
+   * scenario; the cap bounds the reminder, never the detection (the chain key
+   * always compares the FULL canonical string).
+   */
+  argumentsPreviewChars?: number
+}
+
+export const Config: z<Config> = z.object({
+  thresholds: z.array(z.number()).default([3, 5, 8]),
+  include: z.array(z.string()).default([]),
+  exclude: z.array(z.string()).default([]),
+  argumentsPreviewChars: z.number().default(500),
+})
+
+/**
+ * The `{kind:'plugin'}` source stamped on every reminder this guard injects —
+ * the label is load-bearing (an unlabeled context would render as a user
+ * prompt in derived history).
+ */
+const PLUGIN_SOURCE: MessageSource = { kind: 'plugin', plugin: 'repeat-tool-guard' }
+
+/**
+ * The gentle first-threshold reminder. Keyed to `thresholds[0]`, not a literal
+ * count, so a custom first threshold keeps the gentle-then-detailed escalation.
+ */
+const GENTLE_REMINDER =
+  'You are repeating the exact same tool call with identical arguments. '
+  + 'Carefully analyze the previous result before calling again: if the task is '
+  + 'not complete, try a different approach or different arguments instead of '
+  + 'repeating the call.'
+
+/** The detailed later-threshold reminder naming the tool, the run length, and the canonical arguments. */
+function detailedReminder(toolName: string, count: number, canonicalArguments: string): string {
+  return 'Repeated tool call detected:\n'
+    + `- tool: ${toolName}\n`
+    + `- consecutive_calls: ${count}\n`
+    + `- arguments: ${canonicalArguments}\n`
+    + 'The repeated calls are not making progress. Do not call this tool with '
+    + 'these exact arguments again. Inspect the latest result and choose a '
+    + 'different action, different arguments, or finish the task if enough '
+    + 'evidence has been gathered.'
+}
+
+/**
+ * Deep key-sort of a parsed-JSON value so two argument objects that differ
+ * only in property order canonicalize identically. Arguments reach the guard
+ * as the loop's `JSON.parse` output (or its raw-string fallback for malformed
+ * argument JSON), so JSON's value domain is the whole input domain — no
+ * bigint, cycle, or `undefined` handling exists because no input path can
+ * produce them.
+ */
+function sortJsonValue(value: unknown): unknown {
+  if (Array.isArray(value)) return value.map(sortJsonValue)
+  if (value !== null && typeof value === 'object') {
+    const record = value as Record<string, unknown>
+    const sorted: Record<string, unknown> = {}
+    for (const key of Object.keys(record).sort()) {
+      sorted[key] = sortJsonValue(record[key])
+    }
+    return sorted
+  }
+  return value
+}
+
+/** Canonical string form of a call's arguments: deep key-sort, then stringify. */
+function canonicalize(argumentsValue: unknown): string {
+  return JSON.stringify(sortJsonValue(argumentsValue))
+}
+
+/** Compile one `*`-wildcard pattern to an anchored RegExp (every other regex metacharacter is matched literally). */
+function wildcardToRegExp(pattern: string): RegExp {
+  const escaped = pattern.replace(/[|\\{}()[\]^$+?.]/g, String.raw`\$&`)
+  return new RegExp(`^${escaped.replaceAll('*', '.*')}$`)
+}
+
+/**
+ * Head-truncate the canonical arguments for quoting in the detailed reminder,
+ * marking how much was omitted. Bounds only the model-visible text — the
+ * chain key always uses the full canonical string.
+ */
+function previewArguments(canonical: string, cap: number): string {
+  if (canonical.length <= cap) return canonical
+  return `${canonical.slice(0, cap)}… (+${canonical.length - cap} more chars)`
+}
+
+/**
+ * Validate `thresholds` per the fail-loud contract and return them sorted
+ * ascending (the escalation rule reads `thresholds[0]` as the gentle tier, so
+ * order is normalized here, once).
+ */
+function validateThresholds(values: number[]): number[] {
+  if (values.length === 0) {
+    throw new Error('repeat-tool-guard: `thresholds` must not be empty')
+  }
+  for (const value of values) {
+    if (!Number.isInteger(value) || value < 2) {
+      throw new Error(`repeat-tool-guard: invalid threshold ${value} — every threshold must be an integer >= 2`)
+    }
+  }
+  if (new Set(values).size !== values.length) {
+    throw new Error('repeat-tool-guard: `thresholds` must not contain duplicates')
+  }
+  return [...values].sort((a, b) => a - b)
+}
+
+/**
+ * Concatenate the guard's reminder context with a downstream listener's
+ * optional one so folding drops neither. The merged block carries the guard's
+ * `source` — a `HookContext` holds one `MessageSource` and the seam cannot
+ * represent mixed provenance; the rendered `context/message` only
+ * distinguishes by `source.kind`, so a downstream plugin's text is still
+ * correctly framed as plugin context.
+ */
+function concatContext(ours: HookContext, theirs: HookContext | undefined): HookContext {
+  if (!theirs) return ours
+  return { content: [...ours.content, ...theirs.content], source: ours.source }
+}
+
+/** One agent's consecutive-repeat chain: the last tracked call's identity key and its run length. */
+interface Chain {
+  key: string
+  count: number
+}
+
+/**
+ * Install the guard's listeners.
+ * @param ctx - plugin context; listeners are scoped to it and disposed with it.
+ * @param config - validated {@link Config}; `thresholds` is re-checked fail-loud here.
+ */
+export function apply(ctx: Context, config: Config): void {
+  // schemastery's .default() guarantees the fields are set after validation.
+  const thresholds = validateThresholds(config.thresholds as number[])
+  const thresholdSet = new Set(thresholds)
+  const includePatterns = (config.include as string[]).map(wildcardToRegExp)
+  const excludePatterns = (config.exclude as string[]).map(wildcardToRegExp)
+  const argumentsPreviewChars = config.argumentsPreviewChars as number
+  if (!Number.isInteger(argumentsPreviewChars) || argumentsPreviewChars < 1) {
+    throw new Error(`repeat-tool-guard: invalid argumentsPreviewChars ${argumentsPreviewChars} — must be an integer >= 1`)
+  }
+
+  const chains = new Map<AgentId, Chain>()
+
+  /** Whether a tool participates in the chain (untracked calls are transparent: they neither count nor reset). */
+  function tracked(toolName: string): boolean {
+    if (includePatterns.length > 0 && !includePatterns.some(pattern => pattern.test(toolName))) return false
+    return !excludePatterns.some(pattern => pattern.test(toolName))
+  }
+
+  /**
+   * Advance the calling agent's chain for one attempt and return the reminder
+   * to deliver, if this attempt's run length hits a configured threshold.
+   * Counting happens here — in post-execute — because denied calls also flow
+   * through this waterfall (`ToolRegistry.execute` routes a deny through the
+   * same pipeline), and a model hammering a denied call is exactly the loop
+   * worth breaking.
+   */
+  function observe(exec: ToolExecution): HookContext | undefined {
+    // A direct `ctx.tools.execute()` caller has no model to remind and no id
+    // to key on; only agent-loop calls participate.
+    if (!exec.agent) return undefined
+    if (!tracked(exec.name)) return undefined
+    const canonical = canonicalize(exec.arguments)
+    const key = JSON.stringify([exec.name, canonical])
+    const chain = chains.get(exec.agent.id)
+    const count = chain !== undefined && chain.key === key ? chain.count + 1 : 1
+    chains.set(exec.agent.id, { key, count })
+    if (!thresholdSet.has(count)) return undefined
+    const text = count === thresholds[0]
+      ? GENTLE_REMINDER
+      : detailedReminder(exec.name, count, previewArguments(canonical, argumentsPreviewChars))
+    return { content: [{ type: 'text', text }], source: PLUGIN_SOURCE }
+  }
+
+  // Observe-and-enrich, never veto: count first (state advances regardless of
+  // the downstream outcome), DELEGATE so a later listener can still block or
+  // replace, then fold the reminder onto whatever came back — additionalContext
+  // rides both decision variants, so a blocked call still gets the nudge.
+  ctx.on('tools/post-execute', async (exec, _result, next): Promise<PostToolDecision> => {
+    const reminder = observe(exec)
+    const downstream = await next()
+    if (!reminder) return downstream
+    if (downstream.kind === 'block') {
+      return { kind: 'block', feedback: downstream.feedback, additionalContext: concatContext(reminder, downstream.additionalContext) }
+    }
+    return {
+      kind: 'accept',
+      ...downstream.content !== undefined ? { content: downstream.content } : {},
+      additionalContext: concatContext(reminder, downstream.additionalContext),
+    }
+  })
+
+  // A user interjection changes the context; repetition across it is not a
+  // loop. Pure reset hook: always delegates (attaching nothing, vetoing
+  // nothing).
+  ctx.on('agent/prompt-submit', (agent, _content, _source, next): Promise<PromptDecision> => {
+    chains.delete(agent.id)
+    return next()
+  })
+
+  // Drop state when an agent goes away, bounding the map over harness lifetime.
+  ctx.on('agent/status', (agent, status) => {
+    if (status === 'disposed') chains.delete(agent.id)
+  })
+}

packages/guard/repeat-tool-guard/tests/repeat-tool-guard.spec.ts

@@ -0,0 +1,401 @@
+import { describe, expect, it } from 'vitest'
+import { Context } from 'cordis'
+import LlmService, { CallId } from '@deepseek-ai/dsh-llm'
+import SessionStore, { type SessionEvent } from '@deepseek-ai/dsh-session'
+import SystemPrompt from '@deepseek-ai/dsh-system-prompt'
+import ToolRegistry, { defineTool } from '@deepseek-ai/dsh-tools'
+import AgentRegistry, { AgentId } from '@deepseek-ai/dsh-agent'
+import AgentLoop, { type ReactLoopAgent } from '@deepseek-ai/dsh-agent-loop'
+import * as RepeatToolGuard from '@deepseek-ai/dsh-repeat-tool-guard'
+import type { Config } from '@deepseek-ai/dsh-repeat-tool-guard'
+import { MockAdapter, textResponse, toolCallResponse } from '../../../core/agent-loop/tests/mock-adapter.ts'
+
+/**
+ * Behavior suite for the repeat-tool-call guard: chain semantics (identical /
+ * different-tracked / untracked-transparent / per-agent / resets), threshold
+ * escalation incl. the `thresholds[0]` gentle-text rule, canonicalization,
+ * fold-onto-downstream-decision, and fail-loud config validation — all driven
+ * through a real agent loop against a scripted mock adapter (no network).
+ */
+
+/** Boot the core spine + the guard; the caller registers adapters and extra listeners. */
+async function harness(config: Config = {}): Promise<Context> {
+  const ctx = new Context()
+  await ctx.plugin(LlmService)
+  await ctx.plugin(SessionStore)
+  await ctx.plugin(SystemPrompt)
+  await ctx.plugin(ToolRegistry)
+  await ctx.plugin(AgentRegistry)
+  await ctx.plugin(AgentLoop, { agents: [] })
+  await ctx.plugin(RepeatToolGuard, config)
+  ctx.tools.register(defineTool({ name: 'probe', description: 'p', parameters: {}, async execute() { return [{ type: 'text', text: 'ok' }] } }))
+  ctx.tools.register(defineTool({ name: 'other', description: 'o', parameters: {}, async execute() { return [{ type: 'text', text: 'ok' }] } }))
+  return ctx
+}
+
+function waitForIdle(ctx: Context, agent: ReactLoopAgent): Promise<void> {
+  return new Promise((resolve) => { const d = ctx.on('agent/status', (s, st) => { if (s === agent && st === 'idle') { d(); resolve() } }) })
+}
+
+/** Every `context/message` in the agent's log, flattened to joined text + source for terse assertions. */
+function reminders(agent: ReactLoopAgent): { text: string; source: unknown }[] {
+  return [...agent.session.events]
+    .filter((e): e is SessionEvent<'context/message'> => e.type === 'context/message')
+    .map(e => ({
+      text: e.data.content.map(block => block.type === 'text' ? block.text : '').join('|'),
+      source: e.data.source,
+    }))
+}
+
+const GUARD_SOURCE = { kind: 'plugin', plugin: 'repeat-tool-guard' }
+
+describe('threshold escalation', () => {
+  it('reminds gently at the first default threshold (3) and in detail at the second (5)', async () => {
+    const ctx = await harness()
+    const adapter = new MockAdapter([
+      ...Array.from({ length: 5 }, (_, i) => toolCallResponse(`c${i}`, 'probe', { q: 'same' })),
+      textResponse('done'),
+    ])
+    ctx.llm.registerAdapter(['mock'], adapter)
+    const agent = ctx.agentLoop.create(AgentId('a1'), { model: 'mock' })
+    agent.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, agent)
+
+    const found = reminders(agent)
+    expect(found).toHaveLength(2)
+    expect(found[0]!.text).toContain('repeating the exact same tool call')
+    expect(found[0]!.source).toEqual(GUARD_SOURCE)
+    expect(found[1]!.text).toContain('consecutive_calls: 5')
+    expect(found[1]!.text).toContain('- tool: probe')
+    expect(found[1]!.text).toContain('{"q":"same"}')
+    expect(found[1]!.source).toEqual(GUARD_SOURCE)
+  })
+
+  it('keys the gentle text to thresholds[0], not the literal 3', async () => {
+    const ctx = await harness({ thresholds: [4, 2] }) // unsorted on purpose: normalized ascending
+    const adapter = new MockAdapter([
+      ...Array.from({ length: 4 }, (_, i) => toolCallResponse(`c${i}`, 'probe', {})),
+      textResponse('done'),
+    ])
+    ctx.llm.registerAdapter(['mock'], adapter)
+    const agent = ctx.agentLoop.create(AgentId('a1'), { model: 'mock' })
+    agent.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, agent)
+
+    const found = reminders(agent)
+    expect(found).toHaveLength(2)
+    expect(found[0]!.text).toContain('repeating the exact same tool call') // gentle at 2
+    expect(found[1]!.text).toContain('consecutive_calls: 4') // detailed at 4
+  })
+})
+
+describe('chain semantics', () => {
+  it('caps the detailed reminder arguments at argumentsPreviewChars (detection still keys on the full string)', async () => {
+    const ctx = await harness({ thresholds: [2, 3], argumentsPreviewChars: 24 })
+    const bigPayload = 'x'.repeat(400)
+    const adapter = new MockAdapter([
+      toolCallResponse('c1', 'probe', { body: bigPayload }),
+      toolCallResponse('c2', 'probe', { body: bigPayload }),
+      toolCallResponse('c3', 'probe', { body: bigPayload }),
+      textResponse('done'),
+    ])
+    ctx.llm.registerAdapter(['mock'], adapter)
+    const agent = ctx.agentLoop.create(AgentId('a1'), { model: 'mock' })
+    agent.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, agent)
+
+    const found = reminders(agent)
+    expect(found).toHaveLength(2) // gentle at 2, detailed at 3 — full-key matching survived the cap
+    const detailed = found[1]!.text
+    expect(detailed).toContain('- arguments: {"body":"xxxxxxxxxxxxxx') // 24-char head
+    expect(detailed).toContain('… (+387 more chars)')
+    expect(detailed).not.toContain(bigPayload)
+  })
+
+  it('a different tracked call resets the chain', async () => {
+    const ctx = await harness()
+    const adapter = new MockAdapter([
+      toolCallResponse('c1', 'probe', { q: 1 }),
+      toolCallResponse('c2', 'probe', { q: 1 }),
+      toolCallResponse('c3', 'other', {}), // tracked, different → reset
+      toolCallResponse('c4', 'probe', { q: 1 }),
+      toolCallResponse('c5', 'probe', { q: 1 }),
+      toolCallResponse('c6', 'probe', { q: 1 }), // 3rd consecutive AFTER the reset
+      textResponse('done'),
+    ])
+    ctx.llm.registerAdapter(['mock'], adapter)
+    const agent = ctx.agentLoop.create(AgentId('a1'), { model: 'mock' })
+    agent.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, agent)
+
+    expect(reminders(agent)).toHaveLength(1)
+  })
+
+  it('excluded calls are transparent: they neither count nor reset', async () => {
+    const ctx = await harness({ exclude: ['other'] })
+    const adapter = new MockAdapter([
+      toolCallResponse('c1', 'probe', { q: 1 }),
+      toolCallResponse('c2', 'other', {}), // excluded → invisible to the chain
+      toolCallResponse('c3', 'probe', { q: 1 }),
+      toolCallResponse('c4', 'other', {}),
+      toolCallResponse('c5', 'probe', { q: 1 }), // 3rd consecutive probe
+      textResponse('done'),
+    ])
+    ctx.llm.registerAdapter(['mock'], adapter)
+    const agent = ctx.agentLoop.create(AgentId('a1'), { model: 'mock' })
+    agent.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, agent)
+
+    const found = reminders(agent)
+    expect(found).toHaveLength(1)
+    expect(found[0]!.text).toContain('repeating the exact same tool call')
+  })
+
+  it('include patterns track only matching tools (wildcard star)', async () => {
+    const ctx = await harness({ include: ['pro*'] })
+    const adapter = new MockAdapter([
+      toolCallResponse('c1', 'other', {}),
+      toolCallResponse('c2', 'other', {}),
+      toolCallResponse('c3', 'other', {}), // 3 identical, but untracked
+      toolCallResponse('c4', 'probe', {}),
+      toolCallResponse('c5', 'probe', {}),
+      toolCallResponse('c6', 'probe', {}), // 3 identical, tracked
+      textResponse('done'),
+    ])
+    ctx.llm.registerAdapter(['mock'], adapter)
+    const agent = ctx.agentLoop.create(AgentId('a1'), { model: 'mock' })
+    agent.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, agent)
+
+    const found = reminders(agent)
+    expect(found).toHaveLength(1)
+    expect(found[0]!.text).toContain('repeating the exact same tool call')
+  })
+
+  it('escapes regex metacharacters in patterns (a dot matches only a literal dot)', async () => {
+    const ctx = await harness({ exclude: ['pr.be'] }) // would match 'probe' as a regex; must not as a wildcard
+    const adapter = new MockAdapter([
+      ...Array.from({ length: 3 }, (_, i) => toolCallResponse(`c${i}`, 'probe', {})),
+      textResponse('done'),
+    ])
+    ctx.llm.registerAdapter(['mock'], adapter)
+    const agent = ctx.agentLoop.create(AgentId('a1'), { model: 'mock' })
+    agent.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, agent)
+
+    expect(reminders(agent)).toHaveLength(1) // probe was NOT excluded
+  })
+
+  it('canonicalization ignores property order, deeply', async () => {
+    const ctx = await harness()
+    const adapter = new MockAdapter([
+      toolCallResponse('c1', 'probe', { a: 1, nested: { x: [1, 2], y: null } }),
+      toolCallResponse('c2', 'probe', { nested: { y: null, x: [1, 2] }, a: 1 }),
+      toolCallResponse('c3', 'probe', { a: 1, nested: { x: [1, 2], y: null } }),
+      textResponse('done'),
+    ])
+    ctx.llm.registerAdapter(['mock'], adapter)
+    const agent = ctx.agentLoop.create(AgentId('a1'), { model: 'mock' })
+    agent.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, agent)
+
+    expect(reminders(agent)).toHaveLength(1) // all three canonicalize identically
+  })
+
+  it('keys chains per agent: one agent repeating never trips another', async () => {
+    const ctx = await harness()
+    ctx.llm.registerAdapter(['mock-a'], new MockAdapter([
+      toolCallResponse('a1', 'probe', { q: 1 }),
+      toolCallResponse('a2', 'probe', { q: 1 }),
+      textResponse('done'),
+    ]))
+    ctx.llm.registerAdapter(['mock-b'], new MockAdapter([
+      toolCallResponse('b1', 'probe', { q: 1 }),
+      toolCallResponse('b2', 'probe', { q: 1 }),
+      toolCallResponse('b3', 'probe', { q: 1 }),
+      textResponse('done'),
+    ]))
+    const agentA = ctx.agentLoop.create(AgentId('a'), { model: 'mock-a' })
+    const agentB = ctx.agentLoop.create(AgentId('b'), { model: 'mock-b' })
+    agentA.send([{ type: 'text', text: 'go' }])
+    agentB.send([{ type: 'text', text: 'go' }])
+    await Promise.all([waitForIdle(ctx, agentA), waitForIdle(ctx, agentB)])
+
+    expect(reminders(agentA)).toHaveLength(0) // 2 repeats < 3, despite B's 3 in the same registry
+    expect(reminders(agentB)).toHaveLength(1)
+  })
+
+  it('a new user prompt resets the chain', async () => {
+    const ctx = await harness()
+    const adapter = new MockAdapter([
+      toolCallResponse('c1', 'probe', { q: 1 }),
+      toolCallResponse('c2', 'probe', { q: 1 }),
+      textResponse('turn one done'),
+      toolCallResponse('c3', 'probe', { q: 1 }), // without the reset this would be the 3rd
+      textResponse('turn two done'),
+    ])
+    ctx.llm.registerAdapter(['mock'], adapter)
+    const agent = ctx.agentLoop.create(AgentId('a1'), { model: 'mock' })
+    agent.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, agent)
+    agent.send([{ type: 'text', text: 'again' }])
+    await waitForIdle(ctx, agent)
+
+    expect(reminders(agent)).toHaveLength(0)
+  })
+
+  it('drops an agent chain on disposal', async () => {
+    const ctx = await harness({ thresholds: [2] })
+    ctx.llm.registerAdapter(['mock'], new MockAdapter([
+      toolCallResponse('c1', 'probe', { q: 1 }),
+      textResponse('done'),
+      toolCallResponse('c2', 'probe', { q: 1 }), // same id, fresh agent: count 1, not 2
+      textResponse('done'),
+    ]))
+    // Loop agents are torn down by disposing the scope that created them
+    // (the loop.spec pattern): a child plugin fiber owns `first`.
+    let first!: ReactLoopAgent
+    const fiber = await ctx.plugin(Object.assign((inner: Context) => {
+      first = inner.agentLoop.create(AgentId('reused'), { model: 'mock' })
+    }, { inject: ['agentLoop'] }))
+    first.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, first)
+    await fiber.dispose()
+    await first.done
+
+    const second = ctx.agentLoop.create(AgentId('reused'), { model: 'mock' })
+    second.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, second)
+
+    expect(reminders(second)).toHaveLength(0)
+  })
+
+  it('counts denied calls: hammering a denied tool still draws the reminder', async () => {
+    const ctx = await harness({ thresholds: [2] })
+    ctx.on('tools/pre-execute', async () => ({ kind: 'deny' as const, reason: 'sealed' }))
+    const adapter = new MockAdapter([
+      toolCallResponse('c1', 'probe', { q: 1 }),
+      toolCallResponse('c2', 'probe', { q: 1 }),
+      textResponse('done'),
+    ])
+    ctx.llm.registerAdapter(['mock'], adapter)
+    const agent = ctx.agentLoop.create(AgentId('a1'), { model: 'mock' })
+    agent.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, agent)
+
+    expect(reminders(agent)).toHaveLength(1)
+  })
+
+  it('ignores direct executes with no agent (they neither crash nor advance any chain)', async () => {
+    const ctx = await harness({ thresholds: [2] })
+    const direct = await ctx.tools.execute({ callId: CallId('d1'), name: 'probe', arguments: { q: 1 } })
+    expect(direct.isError).toBe(false)
+
+    ctx.llm.registerAdapter(['mock'], new MockAdapter([
+      toolCallResponse('c1', 'probe', { q: 1 }), // if the direct call had counted, this would be #2
+      textResponse('done'),
+    ]))
+    const agent = ctx.agentLoop.create(AgentId('a1'), { model: 'mock' })
+    agent.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, agent)
+
+    expect(reminders(agent)).toHaveLength(0)
+  })
+})
+
+describe('fold onto the downstream decision', () => {
+  it('folds the reminder onto a downstream block and keeps its feedback', async () => {
+    const ctx = await harness({ thresholds: [2] })
+    ctx.on('tools/post-execute', async () => ({
+      kind: 'block' as const,
+      feedback: [{ type: 'text' as const, text: 'nope' }],
+      additionalContext: { content: [{ type: 'text' as const, text: 'downstream-ctx' }], source: { kind: 'plugin' as const, plugin: 'test' } },
+    }))
+    const adapter = new MockAdapter([
+      toolCallResponse('c1', 'probe', { q: 1 }),
+      toolCallResponse('c2', 'probe', { q: 1 }),
+      textResponse('done'),
+    ])
+    ctx.llm.registerAdapter(['mock'], adapter)
+    const agent = ctx.agentLoop.create(AgentId('a1'), { model: 'mock' })
+    agent.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, agent)
+
+    const found = reminders(agent)
+    expect(found).toHaveLength(2)
+    // Call 1: below threshold — the downstream context passes through untouched.
+    expect(found[0]!.text).toBe('downstream-ctx')
+    expect(found[0]!.source).toEqual({ kind: 'plugin', plugin: 'test' })
+    // Call 2: reminder folded in front, single merged context, the guard's source.
+    expect(found[1]!.text).toContain('repeating the exact same tool call')
+    expect(found[1]!.text).toContain('|downstream-ctx')
+    expect(found[1]!.source).toEqual(GUARD_SOURCE)
+    // The block's feedback reached the tool result unchanged.
+    const results = [...agent.session.events].filter((e): e is SessionEvent<'tool/result'> => e.type === 'tool/result')
+    expect(results.every(r => r.data.isError)).toBe(true)
+    expect(results[1]!.data.content).toEqual([{ type: 'text', text: 'nope' }])
+  })
+
+  it('preserves a downstream accept content replacement while folding', async () => {
+    const ctx = await harness({ thresholds: [2] })
+    ctx.on('tools/post-execute', async () => ({
+      kind: 'accept' as const,
+      content: [{ type: 'text' as const, text: 'replaced' }],
+    }))
+    const adapter = new MockAdapter([
+      toolCallResponse('c1', 'probe', { q: 1 }),
+      toolCallResponse('c2', 'probe', { q: 1 }),
+      textResponse('done'),
+    ])
+    ctx.llm.registerAdapter(['mock'], adapter)
+    const agent = ctx.agentLoop.create(AgentId('a1'), { model: 'mock' })
+    agent.send([{ type: 'text', text: 'go' }])
+    await waitForIdle(ctx, agent)
+
+    const found = reminders(agent)
+    expect(found).toHaveLength(1)
+    expect(found[0]!.text).toContain('repeating the exact same tool call')
+    const results = [...agent.session.events].filter((e): e is SessionEvent<'tool/result'> => e.type === 'tool/result')
+    expect(results[1]!.data.content).toEqual([{ type: 'text', text: 'replaced' }])
+  })
+})
+
+describe('config validation fails loud', () => {
+  async function spine(): Promise<Context> {
+    const ctx = new Context()
+    await ctx.plugin(LlmService)
+    await ctx.plugin(SessionStore)
+    await ctx.plugin(SystemPrompt)
+    await ctx.plugin(ToolRegistry)
+    await ctx.plugin(AgentRegistry)
+    await ctx.plugin(AgentLoop, { agents: [] })
+    return ctx
+  }
+
+  it('rejects an empty thresholds list', async () => {
+    const ctx = await spine()
+    await expect(ctx.plugin(RepeatToolGuard, { thresholds: [] })).rejects.toThrow(/must not be empty/)
+  })
+
+  it('rejects a threshold below 2', async () => {
+    const ctx = await spine()
+    await expect(ctx.plugin(RepeatToolGuard, { thresholds: [1, 3] })).rejects.toThrow(/integer >= 2/)
+  })
+
+  it('rejects a non-integer threshold', async () => {
+    const ctx = await spine()
+    await expect(ctx.plugin(RepeatToolGuard, { thresholds: [2.5] })).rejects.toThrow(/integer >= 2/)
+  })
+
+  it('rejects duplicate thresholds', async () => {
+    const ctx = await spine()
+    await expect(ctx.plugin(RepeatToolGuard, { thresholds: [3, 3] })).rejects.toThrow(/duplicates/)
+  })
+
+  it('rejects a non-positive or fractional argumentsPreviewChars', async () => {
+    const ctx = await spine()
+    await expect(ctx.plugin(RepeatToolGuard, { argumentsPreviewChars: 0 })).rejects.toThrow(/argumentsPreviewChars/)
+    const ctx2 = await spine()
+    await expect(ctx2.plugin(RepeatToolGuard, { argumentsPreviewChars: 12.5 })).rejects.toThrow(/argumentsPreviewChars/)
+  })
+})

packages/guard/repeat-tool-guard/tsconfig.json

@@ -0,0 +1,30 @@
+{
+  "extends": "../../../tsconfig.base.json",
+  "compilerOptions": {
+    "rootDir": "src",
+    "outDir": "lib/types"
+  },
+  "include": [
+    "src"
+  ],
+  "references": [
+    {
+      "path": "../../../vendor/cosmokit"
+    },
+    {
+      "path": "../../../vendor/cordis"
+    },
+    {
+      "path": "../../../vendor/schemastery"
+    },
+    {
+      "path": "../../core/tools"
+    },
+    {
+      "path": "../../core/agent"
+    },
+    {
+      "path": "../../llm/llm"
+    }
+  ]
+}

packages/support/acp-snapshot/src/suite.ts

@@ -52,12 +52,22 @@ export interface Scenario {
   /**
    * Whether `test:snapshot:record` regenerates this scenario's `session.jsonl`
    * from the LIVE API. `recorded` scenarios are model-driven and reproducible;
-   * `authored` scenarios (a hand-written `replay.override.json` sidecar drives
-   * replay — e.g. a provider error or a cancel, which the live API can't be
-   * coaxed into deterministically — or a deterministic hook scenario whose
-   * derived empty script needs no sidecar) are NEVER re-recorded.
+   * `authored` scenarios (fixtures hand-written or hand-harvested — e.g. a
+   * provider error or a cancel the live API can't be coaxed into
+   * deterministically, a deterministic hook scenario, or a scripted repetition
+   * a live model won't reproduce) are NEVER re-recorded.
    */
   recorded: boolean
+  /**
+   * Whether replay is driven by a hand-written `replay.override.json` sidecar
+   * (a `ReplayEntry[]` that REPLACES the script derived from `session.jsonl`)
+   * — the throw/hang cases chunks cannot express. The fixture guard requires
+   * the sidecar exactly when this is set: the harness forwards the file purely
+   * on existence, so an unregistered stray sidecar would silently replace the
+   * derived script — the guard fails loud on either mismatch. Defaults to
+   * false (replay derives from the fixture's `assistant/chunk` events).
+   */
+  overridden?: boolean
   /**
    * How many SUBAGENT child sessions this scenario records beyond the top-level
    * one (0 for a single-session scenario). Each child rides in a sibling fixture
@@ -317,17 +327,18 @@ export function defineAcpSnapshotSuite(options: SnapshotSuiteOptions): void {
       // throws "fixture not found" when it is absent and no override replaces it.
       // A no-model scenario ships a header-only `session.jsonl` (it derives to an
       // empty script — no model call is made); a model scenario's fixture also
-      // doubles as the expected-log artifact the run is diffed against. An authored
-      // (non-`recorded`) model scenario additionally ships a `replay.override.json`
-      // sidecar for the throw/hang cases a derived script cannot express.
-      for (const { name, hasModelTurn, recorded, childSessions } of scenarios) {
+      // doubles as the expected-log artifact the run is diffed against. The
+      // `replay.override.json` sidecar is matched BOTH ways against the table's
+      // `overridden` flag: required when set, forbidden when not — the harness
+      // forwards the file purely on existence, so an unregistered stray sidecar
+      // would silently replace the derived script.
+      for (const { name, overridden, childSessions } of scenarios) {
         const dir = join(snapshotsDir, name)
         expect(existsSync(join(dir, 'input.json')), `${name}/input.json`).toBe(true)
         expect(existsSync(join(dir, 'stdout.golden.jsonl')), `${name}/stdout.golden.jsonl`).toBe(true)
         expect(existsSync(join(dir, 'session.jsonl')), `${name}/session.jsonl`).toBe(true)
-        if (hasModelTurn && !recorded) {
-          expect(existsSync(join(dir, 'replay.override.json')), `${name}/replay.override.json`).toBe(true)
-        }
+        expect(existsSync(join(dir, 'replay.override.json')), `${name}/replay.override.json presence must match \`overridden\``)
+          .toBe(overridden === true)
         // A nested-agent scenario ships one child fixture per recorded subagent
         // session (`session.1.jsonl` …), the replay source for that child session.
         for (const childFixture of childFixturePaths(dir, childSessions ?? 0)) {

packages/support/acp-snapshot/tests/suite.spec.ts

@@ -39,14 +39,14 @@ const REPLAY_SCENARIOS: Scenario[] = [
   { name: 'plain-turn', hasModelTurn: true, recorded: true, childSessions: 1 },
   { name: 'no-model', hasModelTurn: false, recorded: false },
   { name: 'blocked-log', hasModelTurn: false, comparesLog: true, recorded: false },
-  { name: 'authored-error', hasModelTurn: true, recorded: false },
+  { name: 'authored-error', hasModelTurn: true, recorded: false, overridden: true },
 ]
 
 const RECORD_SCENARIOS: Scenario[] = [
   { name: 'rec-pin', hasModelTurn: true, recorded: true, pinsHeader: true },
   { name: 'rec-child', hasModelTurn: true, recorded: true, childSessions: 1 },
   // recorded:false in record mode → registered but skipped (never re-recorded).
-  { name: 'rec-skip', hasModelTurn: true, recorded: false },
+  { name: 'rec-skip', hasModelTurn: true, recorded: false, overridden: true },
 ]
 
 // Record mode mutates its snapshots dir, so run it on a throwaway copy —

packages/timeout/README.md

@@ -0,0 +1,9 @@
+# timeout/ — tool-call timeout policy
+
+The tool-call timeout policy plugin. A single **product** package: it is a deployment-policy consumer of the `tools/execute` around-dispatch seam (owned by [`dsh-tools`](../core/tools)) and the pure [`dsh-timeout`](../util/timeout) library — not a swappable capability with an interface/implementation split, so it needs no seam trio.
+
+| Package | Role | ctx key |
+|---|---|---|
+| `timeout-policy/` | A `tools/execute` wrapper: for each configured tool it arms a per-call deadline on `exec.signal` and returns a structured `TOOL_TIMEOUT` result when that deadline wins | (registers a `tools/execute` listener; injects nothing) |
+
+Timeout is split across three layers: [`dsh-timeout`](../util/timeout) owns the pure timing/classification primitive (`deadline`/`timeoutOf`), each capability owns termination (bash kills its process group, the fetch provider tears down its socket), and this package owns the *model-facing tool-call budget as deployment policy* — no model-facing timeout argument, no global default. It is the middleware the [timeout-library RFC](../../docs/rfc/implemented/architecture/2026-07-06-timeout-deadline-library.md) foresaw. `bash` and hook command execution keep their own `BASH_TIMEOUT` backend timeout and do not route through this policy.

packages/timeout/timeout-policy/README.md

@@ -0,0 +1,34 @@
+# dsh-timeout-policy
+
+Tool-call timeout enforcer: a single `tools/execute` around-dispatch listener that arms a per-call cooperative deadline on `exec.signal` for a tool declaring `timeoutMs` on its `ToolDefinition` and returns a structured `TOOL_TIMEOUT` result when that deadline wins. The budget is read from the tool's own declaration (`ToolDefinition.timeoutMs`, set by the owning tool plugin), so this plugin is **zero-config**. It is the reference `tools/execute` wrapper and the enforcement home for model-facing tool-call budgets (the timeout-library RFC's foreseen middleware).
+
+## Plugin (namespace: `timeout-policy`)
+
+A function/namespace plugin (`name` / `inject` / `apply`), not a service. It registers no tool and takes no config — it consumes `ctx.tools`'s `tools/execute` waterfall (which the `dsh-tools` registry always provides) and reads each dispatched tool's declared `timeoutMs` from the registry (`ctx.tools.get(exec.name)`).
+
+```yaml
+- id: timeout-policy
+  name: '@deepseek-ai/dsh-timeout-policy'
+```
+
+The per-tool budget is declared by the tool plugin (e.g. `dsh-tool-web`'s `fetchTimeoutMs`/`searchTimeoutMs` config, attached as `ToolDefinition.timeoutMs`); this plugin only enforces it, so a mistyped tool name is not possible.
+
+### Behavior
+
+For a tool that **declares a `timeoutMs`** the listener:
+
+1. Reads the budget from the tool's own declaration in the registry (`ctx.tools.get(exec.name)?.timeoutMs`) and arms `deadline(exec.signal, timeoutMs, 'TOOL_TIMEOUT')` — one signal fusing the caller's abort with this plugin's timer (`@deepseek-ai/dsh-timeout`).
+2. Swaps that derived signal onto `exec` for the downstream dispatch, then restores the caller's own signal afterward (cordis `next()` ignores passed arguments, so the wrapper mutates the shared `exec` in place; restoring keeps `tools/post-execute` seeing the caller's signal).
+3. After dispatch, if `timeoutOf(d.signal, 'TOOL_TIMEOUT')` matches — this plugin's own timer fired — replaces the result with a structured `TOOL_TIMEOUT` tool result: `{ isError: true, error: { name: 'ToolTimeoutError', code: 'TOOL_TIMEOUT' }, content: 'Error: tool call timed out after <ms>ms' }`.
+
+A tool that **declares no budget** delegates untouched (no deadline).
+
+The base `next()` of `tools/execute` is the registry's dispatch-with-normalization thunk, so when the timeout signal reaches a provider that throws its own upstream-abort error, dispatch first turns it into a normal error result, and this wrapper then replaces that with `TOOL_TIMEOUT`. That ordering is why the replacement is keyed off the signal (`timeoutOf`), not off the dispatched result's shape.
+
+### Cooperative, not a hard kill
+
+The derived signal only **notifies**; termination stays with the tool and the capability it forwards `exec.signal` to (the `dsh-timeout` library owns no kill). **Declaring `timeoutMs` therefore means "cooperative with `exec.signal`"**: a tool that ignores the signal will not stop on timeout. Only signal-forwarding tools should declare it — the shipped `web_fetch`/`web_search` (which forward through `ctx.web` to providers) are the reference. `TOOL_TIMEOUT` needs no session event for reconstructability: it is the final model-facing `tool/result`, already logged by the loop.
+
+### Composing with other `tools/execute` wrappers
+
+Multiple `tools/execute` listeners compose by cordis registration order. Combined with a future retry/sandbox/metrics wrapper, registration order chooses the semantics — "timeout covers the whole retry operation" (timeout registered outer) versus "timeout covers each attempt" (timeout registered inner).

packages/timeout/timeout-policy/package.json

@@ -0,0 +1,36 @@
+{
+  "name": "@deepseek-ai/dsh-timeout-policy",
+  "description": "Tool-call timeout policy: a tools/execute wrapper that arms a per-tool deadline on exec.signal and returns TOOL_TIMEOUT when it wins",
+  "version": "0.0.1",
+  "private": true,
+  "type": "module",
+  "main": "lib/index.js",
+  "types": "lib/types/index.d.ts",
+  "exports": {
+    ".": {
+      "types": "./lib/types/index.d.ts",
+      "default": "./lib/index.js"
+    },
+    "./src/*": "./src/*",
+    "./package.json": "./package.json"
+  },
+  "files": [
+    "lib/index.js",
+    "lib/types/**/*.d.ts",
+    "lib/types/**/*.d.ts.map",
+    "src"
+  ],
+  "license": "BSD-3-Clause",
+  "peerDependencies": {
+    "@deepseek-ai/dsh-llm": "^0.0.1",
+    "@deepseek-ai/dsh-timeout": "^0.0.1",
+    "@deepseek-ai/dsh-tools": "^0.0.1",
+    "cordis": "^4.0.0-rc.6"
+  },
+  "devDependencies": {
+    "@deepseek-ai/dsh-llm": "workspace:^",
+    "@deepseek-ai/dsh-timeout": "workspace:^",
+    "@deepseek-ai/dsh-tools": "workspace:^",
+    "cordis": "^4.0.0-rc.6"
+  }
+}

packages/timeout/timeout-policy/src/index.ts

@@ -0,0 +1,115 @@
+/**
+ * `@deepseek-ai/dsh-timeout-policy`: the tool-call timeout ENFORCER. It registers
+ * ONE `tools/execute` around-dispatch listener that, for a tool declaring a
+ * `timeoutMs` on its {@link ToolDefinition}, arms a per-call deadline on
+ * `exec.signal` and returns a structured `TOOL_TIMEOUT` result when that deadline
+ * wins. The budget is DECLARED by the tool (see `ToolDefinition.timeoutMs`, set
+ * by the owning tool plugin from its own config); this plugin only enforces it,
+ * so it is zero-config and there is no tool-name map to mistype.
+ *
+ * This is a COOPERATIVE deadline, not a hard kill: the derived signal only
+ * NOTIFIES. A tool that declares `timeoutMs` (and the capability it forwards
+ * `exec.signal` to) must honor that signal and reach quiescence — the plugin
+ * never races the tool promise or terminates work itself (see the timeout-library
+ * RFC's rejection of `Promise.race`). Declaring `timeoutMs` therefore MEANS "this
+ * tool is cooperative with `exec.signal`": a tool that ignores the signal will
+ * not stop on timeout, so only signal-forwarding tools should declare it (the
+ * shipped web tools are the reference).
+ *
+ * Ownership of the `TOOL_TIMEOUT` code is entirely here: it is both the internal
+ * {@link deadline} code (so {@link timeoutOf} scopes the classification to THIS
+ * plugin's own timer, reading a foreign/nested outer deadline as an ordinary
+ * cancel) and the structured `{ name, code }` on the replacement tool result.
+ * No new session event is needed for reconstructability: the `TOOL_TIMEOUT`
+ * result IS the final model-facing `tool/result`, already logged by the loop.
+ *
+ * Why a `tools/execute` around seam and not a `pre`/`post` pair: the deadline
+ * needs ONE lexical scope — arm on `exec.signal`, delegate to dispatch, classify
+ * the result, dispose the timer — which the around seam gives directly. A
+ * pre/post split would spread one deadline's lifetime across two independent
+ * waterfalls (a call-id map, cleanup on every deny/throw/dispose path).
+ *
+ * @module @deepseek-ai/dsh-timeout-policy
+ */
+
+import type { Context } from 'cordis'
+import type { CallId } from '@deepseek-ai/dsh-llm'
+import { deadline, timeoutOf } from '@deepseek-ai/dsh-timeout'
+import type { ToolExecutionResult } from '@deepseek-ai/dsh-tools'
+
+/**
+ * The code owned by this plugin, used BOTH as the internal {@link deadline}
+ * classification code AND as the structured error `code` on the replacement
+ * tool result. Scoping {@link timeoutOf} to it keeps a nested outer deadline
+ * (another `tools/execute` wrapper's timer that fired first) from being misread
+ * as this plugin's own timeout — it reads as an ordinary upstream cancel.
+ */
+export const TOOL_TIMEOUT = 'TOOL_TIMEOUT'
+
+/** Cordis plugin name used by loader diagnostics. */
+export const name = 'timeout-policy'
+
+/** The tool registry seam this plugin wraps (`tools/execute`) and reads (`get`). */
+export const inject = ['tools']
+
+/**
+ * The structured result substituted when this plugin's deadline wins. `content`
+ * is the model-facing message; `error.code` is the same {@link TOOL_TIMEOUT}
+ * this plugin owns, so a retry/sandbox plugin (and replay) can route on it.
+ *
+ * @param callId - the timed-out call's id, carried onto the replacement result.
+ * @param timeoutMs - the elapsed budget, rendered into the model-facing message.
+ * @returns the `isError` {@link ToolExecutionResult} with a `TOOL_TIMEOUT` error.
+ */
+export function toolTimeoutResult(callId: CallId, timeoutMs: number): ToolExecutionResult {
+  return {
+    callId,
+    content: [{ type: 'text', text: `Error: tool call timed out after ${timeoutMs}ms` }],
+    isError: true,
+    error: { name: 'ToolTimeoutError', code: TOOL_TIMEOUT },
+  }
+}
+
+/**
+ * Register the tool-call timeout enforcer. For a tool whose {@link ToolDefinition}
+ * declares `timeoutMs`, the listener arms a {@link deadline} on the caller's
+ * `exec.signal`, swaps it onto `exec` for the downstream dispatch (cordis
+ * `next()` ignores passed arguments, so a wrapper mutates the shared `exec` in
+ * place), restores the original signal afterward so `tools/post-execute` sees the
+ * caller's own signal, and replaces the result with {@link toolTimeoutResult}
+ * when its own timer fired. A tool that declares no budget delegates untouched.
+ *
+ * The budget source is the tool's own declaration read from the registry
+ * (`ctx.tools.get(exec.name)?.timeoutMs`), NOT a plugin config map — `exec.name`
+ * is the tool being dispatched, so the lookup always resolves and there is no
+ * mistypable tool name and no unknown-name path to warn or throw about.
+ */
+export function apply(ctx: Context): void {
+  ctx.on('tools/execute', async (exec, next): Promise<ToolExecutionResult> => {
+    const timeoutMs = ctx.tools.get(exec.name)?.timeoutMs
+    // A tool that declares no budget: no deadline, delegate unchanged.
+    if (timeoutMs === undefined) return next()
+
+    using d = deadline(exec.signal, timeoutMs, TOOL_TIMEOUT)
+    // Swap the derived deadline onto exec for dispatch, then restore the
+    // caller's own signal so post-execute listeners never see this plugin's
+    // (possibly already-aborted) timeout signal. `undefined` is not assignable to
+    // the optional `signal` under exactOptionalPropertyTypes, so branch on it.
+    const upstream = exec.signal
+    exec.signal = d.signal
+    try {
+      const result = await next()
+      // If OUR timer fired (scoped by code — a nested outer deadline reads as
+      // undefined here), the tool/capability saw the abort and reached
+      // quiescence; replace whatever it returned (its own abort result) with the
+      // structured TOOL_TIMEOUT the model sees.
+      if (timeoutOf(d.signal, TOOL_TIMEOUT) !== undefined) {
+        return toolTimeoutResult(exec.callId, timeoutMs)
+      }
+      return result
+    } finally {
+      if (upstream === undefined) delete exec.signal
+      else exec.signal = upstream
+    }
+  })
+}

packages/timeout/timeout-policy/tests/timeout-policy.spec.ts

@@ -0,0 +1,200 @@
+/**
+ * Unit + real-load-path coverage for @deepseek-ai/dsh-timeout-policy. The
+ * timeout-wins cases drive the deadline under fake timers (deterministic — no
+ * wall-clock race) and use a COOPERATIVE tool that settles only when its
+ * `exec.signal` aborts, mirroring how a real capability forwards the signal and
+ * reaches quiescence.
+ */
+
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
+import { Context } from 'cordis'
+import Loader from '@cordisjs/plugin-loader'
+import { CallId, HarnessError } from '@deepseek-ai/dsh-llm'
+import SystemPrompt from '@deepseek-ai/dsh-system-prompt'
+import ToolRegistry, { defineTool, type ToolExecution, type ToolExecutionResult, type PostToolDecision } from '@deepseek-ai/dsh-tools'
+import * as timeoutPolicy from '@deepseek-ai/dsh-timeout-policy'
+import { TOOL_TIMEOUT, toolTimeoutResult } from '@deepseek-ai/dsh-timeout-policy'
+
+/** Mount the registry + the zero-config timeout-policy enforcer. */
+async function setup() {
+  const ctx = new Context()
+  await ctx.plugin(SystemPrompt)
+  await ctx.plugin(ToolRegistry)
+  await ctx.plugin(timeoutPolicy)
+  return ctx
+}
+
+/** A cooperative tool that settles ONLY when its exec.signal aborts (returns text). */
+const cooperativeTool = defineTool({
+  name: 'slow', description: 'stops when aborted', parameters: {}, timeoutMs: 100,
+  execute(_args, exec): Promise<{ type: 'text'; text: string }[]> {
+    const done = [{ type: 'text' as const, text: 'stopped cooperatively' }]
+    if (exec.signal?.aborted) return Promise.resolve(done)
+    return new Promise((resolve) => { exec.signal?.addEventListener('abort', () => { resolve(done) }) })
+  },
+})
+
+/** A cooperative tool that THROWS its own upstream-abort error when aborted (web-provider shape). */
+const abortThrowingTool = defineTool({
+  name: 'aborter', description: 'throws WEB_ABORTED when aborted', parameters: {}, timeoutMs: 100,
+  execute(_args, exec): Promise<never> {
+    if (exec.signal?.aborted) return Promise.reject(new HarnessError('web fetch aborted', 'WEB_ABORTED'))
+    return new Promise((_resolve, reject) => { exec.signal?.addEventListener('abort', () => { reject(new HarnessError('web fetch aborted', 'WEB_ABORTED')) }) })
+  },
+})
+
+describe('timeout-policy delegation (unconfigured / fast)', () => {
+  it('delegates a tool with NO declared budget unchanged and does not touch exec.signal', async () => {
+    const ctx = await setup()
+    let seenSignal: AbortSignal | undefined
+    ctx.tools.register(defineTool({ name: 'probe', description: 'd', parameters: {},
+      async execute(_a, exec) { seenSignal = exec.signal; return [{ type: 'text' as const, text: 'ok' }] } }))
+    const upstream = new AbortController().signal
+    const result = await ctx.tools.execute({ callId: CallId('c1'), name: 'probe', arguments: {}, signal: upstream })
+    expect(result.isError).toBe(false)
+    expect(seenSignal).toBe(upstream)
+  })
+
+  it('a tool with a budget that returns fast keeps its own result (no timeout)', async () => {
+    const ctx = await setup()
+    ctx.tools.register(defineTool({ name: 'fast', description: 'd', parameters: {}, timeoutMs: 10_000,
+      async execute() { return [{ type: 'text' as const, text: 'ok' }] } }))
+    const result = await ctx.tools.execute({ callId: CallId('c1'), name: 'fast', arguments: {} })
+    expect(result).toEqual({ callId: CallId('c1'), content: [{ type: 'text', text: 'ok' }], isError: false })
+  })
+
+  it('a budgeted tool receives the DERIVED deadline signal (not the caller signal) during dispatch', async () => {
+    const ctx = await setup()
+    let seenSignal: AbortSignal | undefined
+    ctx.tools.register(defineTool({ name: 'probe', description: 'd', parameters: {}, timeoutMs: 10_000,
+      async execute(_a, exec) { seenSignal = exec.signal; return [{ type: 'text' as const, text: 'ok' }] } }))
+    const upstream = new AbortController().signal
+    await ctx.tools.execute({ callId: CallId('c1'), name: 'probe', arguments: {}, signal: upstream })
+    expect(seenSignal).toBeDefined()
+    expect(seenSignal).not.toBe(upstream)
+  })
+})
+
+describe('timeout-policy signal restoration', () => {
+  it('restores the caller signal for post-execute after wrapping', async () => {
+    const ctx = await setup()
+    ctx.tools.register(defineTool({ name: 'fast', description: 'd', parameters: {}, timeoutMs: 10_000,
+      async execute() { return [{ type: 'text' as const, text: 'ok' }] } }))
+    let postSignal: AbortSignal | undefined | 'unset' = 'unset'
+    ctx.on('tools/post-execute', async (exec, _result, next): Promise<PostToolDecision> => { postSignal = exec.signal; return next() })
+    const upstream = new AbortController().signal
+    await ctx.tools.execute({ callId: CallId('c1'), name: 'fast', arguments: {}, signal: upstream })
+    expect(postSignal).toBe(upstream)
+  })
+
+  it('deletes exec.signal again when the caller passed none', async () => {
+    const ctx = await setup()
+    ctx.tools.register(defineTool({ name: 'fast', description: 'd', parameters: {}, timeoutMs: 10_000,
+      async execute() { return [{ type: 'text' as const, text: 'ok' }] } }))
+    let hadSignal: boolean | undefined
+    ctx.on('tools/post-execute', async (exec, _result, next): Promise<PostToolDecision> => { hadSignal = 'signal' in exec && exec.signal !== undefined; return next() })
+    await ctx.tools.execute({ callId: CallId('c1'), name: 'fast', arguments: {} })
+    expect(hadSignal).toBe(false)
+  })
+})
+
+describe('timeout-policy TOOL_TIMEOUT replacement (deadline wins)', () => {
+  beforeEach(() => { vi.useFakeTimers() })
+  afterEach(() => { vi.useRealTimers() })
+
+  it('replaces a cooperative tool result with TOOL_TIMEOUT when its own deadline fires', async () => {
+    const ctx = await setup()
+    ctx.tools.register(cooperativeTool)
+    const pending = ctx.tools.execute({ callId: CallId('c1'), name: 'slow', arguments: {} })
+    await vi.advanceTimersByTimeAsync(150)
+    const result = await pending
+    expect(result).toEqual({
+      callId: CallId('c1'),
+      content: [{ type: 'text', text: 'Error: tool call timed out after 100ms' }],
+      isError: true,
+      error: { name: 'ToolTimeoutError', code: 'TOOL_TIMEOUT' },
+    })
+  })
+
+  it('replaces a provider-owned abort ERROR result with TOOL_TIMEOUT when the signal was ours', async () => {
+    const ctx = await setup()
+    ctx.tools.register(abortThrowingTool)
+    const pending = ctx.tools.execute({ callId: CallId('c1'), name: 'aborter', arguments: {} })
+    await vi.advanceTimersByTimeAsync(150)
+    const result = await pending
+    expect(result.isError).toBe(true)
+    expect(result.error).toEqual({ name: 'ToolTimeoutError', code: 'TOOL_TIMEOUT' })
+    expect(result.content[0]).toMatchObject({ text: 'Error: tool call timed out after 100ms' })
+  })
+
+  it('does NOT replace when the caller aborts first (upstream cancel, not our timeout)', async () => {
+    const ctx = await setup()
+    ctx.tools.register(cooperativeTool)
+    const upstream = new AbortController()
+    const pending = ctx.tools.execute({ callId: CallId('c1'), name: 'slow', arguments: {}, signal: upstream.signal })
+    upstream.abort('user cancelled')
+    await vi.advanceTimersByTimeAsync(0)
+    const result = await pending
+    expect(result.isError).toBe(false)
+    expect(result.content[0]).toMatchObject({ text: 'stopped cooperatively' })
+  })
+})
+
+describe('toolTimeoutResult', () => {
+  it('builds the structured TOOL_TIMEOUT result', () => {
+    expect(toolTimeoutResult(CallId('c9'), 250)).toEqual({
+      callId: CallId('c9'),
+      content: [{ type: 'text', text: 'Error: tool call timed out after 250ms' }],
+      isError: true,
+      error: { name: 'ToolTimeoutError', code: 'TOOL_TIMEOUT' },
+    } satisfies ToolExecutionResult)
+  })
+
+  it('exposes the owned code constant', () => {
+    expect(TOOL_TIMEOUT).toBe('TOOL_TIMEOUT')
+  })
+})
+
+describe('timeout-policy disposal (HMR safety)', () => {
+  it('removes its tools/execute listener when the plugin fiber disposes', async () => {
+    const ctx = new Context()
+    await ctx.plugin(SystemPrompt)
+    await ctx.plugin(ToolRegistry)
+    let seenSignal: AbortSignal | undefined
+    ctx.tools.register(defineTool({ name: 'probe', description: 'd', parameters: {}, timeoutMs: 10_000,
+      async execute(_a, exec) { seenSignal = exec.signal; return [{ type: 'text' as const, text: 'ok' }] } }))
+    const fiber = await ctx.plugin(timeoutPolicy)
+    const upstream = new AbortController().signal
+    await ctx.tools.execute({ callId: CallId('c1'), name: 'probe', arguments: {}, signal: upstream })
+    expect(seenSignal).not.toBe(upstream)
+    await fiber.dispose()
+    await ctx.tools.execute({ callId: CallId('c2'), name: 'probe', arguments: {}, signal: upstream })
+    expect(seenSignal).toBe(upstream)
+  })
+})
+
+describe('dsh-timeout-policy real-load-path guard', () => {
+  it('has no default export and keeps name/inject through unwrapExports', () => {
+    expect('default' in timeoutPolicy).toBe(false)
+    const loader = Object.create(Loader.prototype) as Loader
+    const unwrapped = loader.unwrapExports(timeoutPolicy) as Record<string, unknown>
+    expect(unwrapped).toBe(timeoutPolicy)
+    expect(unwrapped.name).toBe('timeout-policy')
+    expect(unwrapped.inject).toEqual(['tools'])
+    expect(typeof unwrapped.apply).toBe('function')
+  })
+
+  it('boots over ctx.tools through the unwrapped module and wraps a budgeted tool', async () => {
+    const ctx = new Context()
+    await ctx.plugin(SystemPrompt)
+    await ctx.plugin(ToolRegistry)
+    ctx.tools.register(defineTool({ name: 'fast', description: 'd', parameters: {}, timeoutMs: 5_000,
+      async execute() { return [{ type: 'text' as const, text: 'ok' }] } }))
+    const loader = Object.create(Loader.prototype) as Loader
+    const unwrapped = loader.unwrapExports(timeoutPolicy) as Parameters<Context['plugin']>[0]
+    const fiber = await ctx.plugin(unwrapped)
+    const result = await ctx.tools.execute({ callId: CallId('c1'), name: 'fast', arguments: {} } satisfies ToolExecution)
+    expect(result.isError).toBe(false)
+    await fiber.dispose()
+  })
+})

packages/timeout/timeout-policy/tsconfig.json

@@ -0,0 +1,16 @@
+{
+  "extends": "../../../tsconfig.base.json",
+  "compilerOptions": {
+    "rootDir": "src",
+    "outDir": "lib/types"
+  },
+  "include": ["src"],
+  "references": [
+    { "path": "../../../vendor/cosmokit" },
+    { "path": "../../../vendor/cordis" },
+    { "path": "../../../vendor/schemastery" },
+    { "path": "../../llm/llm" },
+    { "path": "../../util/timeout" },
+    { "path": "../../core/tools" }
+  ]
+}

packages/ui/README.md

@@ -5,10 +5,14 @@ Integrations that expose the agent to an external editor or client. These are **
 | Package | Role | ctx key |
 |---|---|---|
 | `acp/` | Agent Client Protocol bridge: serves the agent to an ACP editor (Zed) over JSON-RPC stdio | (drives `ctx.agents`/`ctx.sessions`) |
+| `user-interaction/` | Abstract human question/answer seam used by UI-backed confirmation tools | `ctx.userInteraction` |
+| `tool-ask-user/` | Model-facing `ask_user_question` tool over `ctx.userInteraction` | (registers on `ctx.tools`) |
 | `stdio-agent/` | Terminal stdio chat APP: the agent-core spine + console logger + readline UI + a pre-created `main` agent, with a `bin` | (composition + `bin`) |
 | `acp-agent/` | ACP server APP: the agent-core spine + JSONL persistence + the `acp` bridge (no stdout logger), with a `bin` | (composition + `bin`) |
 | `app-boot/` | Shared boot glue for the two app bins: `.env` loading, fail-loud Loader guards, snapshot-aware config resolution, the settle-the-tree boot sequence | (library for the bins) |
 
 A UI integration is a client-driver plugin, not a loop change and not a capability seam: it consumes the existing `agent/*` event taxonomy and the `dsh-agent` factory. The readline UI is the unstructured analogue of the `acp` bridge and lives INSIDE the stdio app (the `stdio-chat` module of [`stdio-agent/`](stdio-agent/README.md)): it is scaffolding for that one front door, not an independently swappable integration, so it carries no package boundary of its own.
 
+`user-interaction` and `tool-ask-user` live here because asking a human is a UI-backed product affordance, not part of the providerless core spine. The seam remains provider-neutral (`ctx.userInteraction`), while the tool is the model-facing consumer and the app/bridge packages provide concrete providers.
+
 `stdio-agent` and `acp-agent` are the two **app packages**: each composes the [`core/agent-core`](../core/agent-core/README.md) spine with its coupled front-door cluster (and owns the boot `bin`), so a leaf `cordis.yml` is the swappable backends plus one app entry plus any optional product tools. They live in `ui/` because each IS a user-facing front door; the stdout-purity coupling (logger vs. no logger) becomes a property of the artifact rather than a leaf convention.

packages/ui/acp-agent/README.md

@@ -11,8 +11,10 @@ stdout is the ACP JSON-RPC channel, so the cluster is defined as much by what it
 | Plugin | Why |
 |---|---|
 | `@deepseek-ai/dsh-agent-core` | the spine, pre-creating **no** agents (ACP `session/new` creates them on demand) |
+| `@deepseek-ai/dsh-user-interaction` | the human question/answer seam used by clients that can complete ACP elicitation requests |
 | `@deepseek-ai/dsh-session-persistence-jsonl` | durable JSONL session log (the bridge advertises `loadSession`) |
-| `@deepseek-ai/dsh-acp` | the bridge that owns stdout for JSON-RPC |
+| `@deepseek-ai/dsh-acp` | the bridge that owns stdout for JSON-RPC and provides ACP-backed user answers when a leaf explicitly exposes a user-question tool |
+| ~~`@deepseek-ai/dsh-tool-ask-user`~~ | **omitted by default** — ACP elicitation support is still client-dependent, so leaves must opt in deliberately |
 | ~~console logger~~ | **omitted** — it writes to stdout and would corrupt the protocol frames ([the stdout-purity footgun](../acp/README.md)) |
 | ~~`hmr`~~ | **omitted** — the editor owns the subprocess |
 

packages/ui/acp-agent/package.json

@@ -36,6 +36,7 @@
     "@deepseek-ai/dsh-acp": "^0.0.1",
     "@deepseek-ai/dsh-agent-core": "^0.0.1",
     "@deepseek-ai/dsh-session-persistence-jsonl": "^0.0.1",
+    "@deepseek-ai/dsh-user-interaction": "^0.0.1",
     "cordis": "^4.0.0-rc.6",
     "schemastery": "^3.17.0"
   },
@@ -47,6 +48,7 @@
     "@deepseek-ai/dsh-agent-core": "workspace:^",
     "@deepseek-ai/dsh-system-prompt": "workspace:^",
     "@deepseek-ai/dsh-session-persistence-jsonl": "workspace:^",
+    "@deepseek-ai/dsh-user-interaction": "workspace:^",
     "cordis": "^4.0.0-rc.6",
     "schemastery": "^3.17.0"
   }

packages/ui/acp-agent/src/index.ts

@@ -35,6 +35,7 @@ import z from 'schemastery'
 import * as acp from '@deepseek-ai/dsh-acp'
 import * as agentCore from '@deepseek-ai/dsh-agent-core'
 import SessionPersistenceJsonl from '@deepseek-ai/dsh-session-persistence-jsonl'
+import UserInteractionService from '@deepseek-ai/dsh-user-interaction'
 
 export const name = 'acp-agent'
 
@@ -79,6 +80,7 @@ export function apply(ctx: Context, config: Config): void {
     ...config.persona !== undefined ? { persona: config.persona } : {},
     ...config.toolOrder !== undefined ? { toolOrder: config.toolOrder } : {},
   })
+  ctx.plugin(UserInteractionService)
   ctx.plugin(SessionPersistenceJsonl, { root: config.persistenceRoot ?? './.sessions' })
   ctx.plugin(acp, { model: config.model })
 }

packages/ui/acp-agent/tests/acp-agent.spec.ts

@@ -30,6 +30,8 @@ describe('dsh-acp-agent composition', () => {
     expect(ctx.get('sessions')).toBeDefined()
     expect(ctx.get('sessionPersistence')).toBeDefined()
     expect(ctx.get('agentLoop')).toBeDefined()
+    expect(ctx.get('userInteraction')).toBeDefined()
+    expect(ctx.get('tools')?.get('ask_user_question')).toBeUndefined()
     // No pre-created agents — ACP session/new creates them on demand.
     expect(ctx.get('agents')!.list()).toHaveLength(0)
     await ctx.fiber.dispose()

packages/ui/acp-agent/tsconfig.json

@@ -26,6 +26,12 @@
     {
       "path": "../../core/agent-core"
     },
+    {
+      "path": "../user-interaction"
+    },
+    {
+      "path": "../tool-ask-user"
+    },
     {
       "path": "../../session-persistence/session-persistence-jsonl"
     }

packages/ui/acp/README.md

@@ -8,7 +8,7 @@ It is a **client-driver / UI plugin**, the structured analogue of the readline `
 
 `apply(ctx, config)` — wires an `AgentSideConnection` (from `@agentclientprotocol/sdk`) to `process.stdin`/`process.stdout` and implements the ACP `Agent` method surface.
 
-`inject: ['agents', 'sessions', 'sessionPersistence', 'tools']` — programs against the interface packages only (never `dsh-agent-loop`). `sessionPersistence` is required because `initialize` advertises `loadSession: true`; `tools` lets a tool own how its calls render (`presentCall`/`presentResult`) — the bridge looks the definition up by name and falls back to a generic presentation when a tool declares none (see Tool-call presentation).
+`inject: ['agents', 'sessions', 'sessionPersistence', 'tools', 'userInteraction']` — programs against the interface packages only (never `dsh-agent-loop`). `sessionPersistence` is required because `initialize` advertises `loadSession: true`; `tools` lets a tool own how its calls render (`presentCall`/`presentResult`) — the bridge looks the definition up by name and falls back to a generic presentation when a tool declares none (see Tool-call presentation). `userInteraction` lets agent-owned `ask_user_question` calls become ACP form elicitations routed to the owning session.
 
 ### Config
 
@@ -30,6 +30,7 @@ The `initialize` handshake reports a fixed server identity (`agentInfo: { name:
 | `session/prompt` | `agent.send()` | supports ACP `text` and `resource_link` blocks; rejects image/audio/embedded resource and empty prompts; one in-flight prompt PER session (independent); settles on the OWNING turn's end (a turn that ends in `error` rejects the RPC) |
 | `session/cancel` | `agent.cancel()` | the queue-aware cancel: aborts a running step, clears queued + steering work, and drops a turn about to start, then settles the prompt `cancelled` — for ONLY that session (a cancel never touches another session's stream or prompt) |
 | `session/update` | `session/event` | `agent_message_chunk` (text-delta), `agent_thought_chunk` (reasoning-delta), `user_message_chunk` (load replay), `tool_call`/`tool_call_update` (the render intent — a `card`-tagged `ToolCallView`/`ToolResultView` — owned by the TOOL via `presentCall`/`presentResult`, which the bridge switches on to build the wire shape — see Tool-call presentation) |
+| `elicitation/create` | `ctx.userInteraction.ask()` | maps `ask_user_question` questions to ACP form elicitations; option descriptions are shown in enum titles, `multi_select` uses ACP array enums, optionless requests use a required `custom` field, and a non-empty custom answer overrides any selected choice |
 
 ## Multi-session
 

packages/ui/acp/acp-feature-support.md

@@ -47,7 +47,7 @@ These are capabilities the bridge would *drive* on the editor. The harness runs
 | `terminal/wait_for_exit` | S | ❌ | ❌ | ❌ | As above. |
 | `terminal/kill` | S | ❌ | ❌ | ❌ | As above. |
 | `terminal/release` | S | ❌ | ❌ | ❌ | As above. |
-| `elicitation/create` · `elicitation/complete` | U | ❌ | ✅ | ⚠️ | Structured user-input forms. Claude calls the `unstable_*` elicitation methods (to surface MCP server elicitations); Codex does NOT — its `CodexElicitationHandler` maps elicitations onto `session/request_permission` instead. |
+| `elicitation/create` · `elicitation/complete` | U | ⚠️ | ✅ | ⚠️ | The bridge drives `unstable_createElicitation` for `ask_user_question` form prompts (session-scoped, no URL-mode flow yet). Claude calls the `unstable_*` elicitation methods for MCP server elicitations; Codex maps elicitations onto `session/request_permission`. |
 
 ## 3. Capabilities
 
@@ -140,7 +140,7 @@ The bridge rejects unsupported prompt blocks rather than silently dropping them
 
 Ranked by how commonly the reference adapters ship them and how much UX they unlock:
 
-1. **Permission gate** — `session/request_permission` + permission options. Tracked `TODO(rfc010-permission-gate)`; the reverse map is already wired. Foundational, and a prerequisite for modes.
+1. **Permission gate** — `session/request_permission` + permission options. Tracked `TODO(rfc010-permission-gate)`; the reverse map is already wired and shared with `ask_user_question` routing. Foundational, and a prerequisite for modes.
 2. **Session lifecycle** — `session/list` + `session/delete` (the persistence layer already lists), then `session/resume` / `session/close`.
 3. **Modes / config options / model selection** — coupled to the permission gate.
 4. **Agent plan** (`sessionUpdate: 'plan'`) — surface the loop's plan as structured entries.

packages/ui/acp/package.json

@@ -32,6 +32,7 @@
     "@deepseek-ai/dsh-session": "^0.0.1",
     "@deepseek-ai/dsh-session-persistence": "^0.0.1",
     "@deepseek-ai/dsh-tools": "^0.0.1",
+    "@deepseek-ai/dsh-user-interaction": "^0.0.1",
     "cordis": "^4.0.0-rc.6"
   },
   "devDependencies": {
@@ -49,6 +50,8 @@
     "@deepseek-ai/dsh-tool-fs": "workspace:^",
     "@deepseek-ai/dsh-tool-todo": "workspace:^",
     "@deepseek-ai/dsh-tools": "workspace:^",
+    "@deepseek-ai/dsh-tool-ask-user": "workspace:^",
+    "@deepseek-ai/dsh-user-interaction": "workspace:^",
     "cordis": "^4.0.0-rc.6"
   }
 }

packages/ui/acp/src/index.ts

@@ -47,6 +47,9 @@ import {
   type AuthenticateRequest,
   type CancelNotification,
   type ContentBlock as AcpContentBlock,
+  type CreateElicitationRequest,
+  type ElicitationContentValue,
+  type EnumOption,
   type InitializeRequest,
   type InitializeResponse,
   type LoadSessionRequest,
@@ -71,6 +74,14 @@ import type { ToolCallView, ToolRegistry, ToolResultView, TerminalResultView } f
 // Side-effect type import: declaration-merges `ctx.sessionPersistence` onto
 // Context (the bridge injects it and reads `list()` for load cwd validation).
 import type {} from '@deepseek-ai/dsh-session-persistence'
+import {
+  UserInteractionError,
+  type AskUserQuestionAnswer,
+  type AskUserQuestionAnswerItem,
+  type AskUserQuestionItem,
+  type AskUserQuestionOption,
+  type AskUserQuestionRequest,
+} from '@deepseek-ai/dsh-user-interaction'
 import {
   acpPromptToText,
   harnessBlockToAcpContent,
@@ -84,7 +95,7 @@ export const name = 'acp'
 // because `initialize` advertises `loadSession: true`. `tools` lets a tool own
 // how its calls render (`presentCall`/`presentResult`); the bridge looks up the
 // definition by name and falls back to a generic presentation when absent.
-export const inject = ['agents', 'sessions', 'sessionPersistence', 'tools']
+export const inject = ['agents', 'sessions', 'sessionPersistence', 'tools', 'userInteraction']
 
 /**
  * Build an ACP "invalid params" error whose human detail rides in the message.
@@ -111,6 +122,116 @@ function sameWorkspaceCwd(left: string, right: string): boolean {
   return resolvePath(left) === resolvePath(right)
 }
 
+function optionDescription(option: AskUserQuestionOption): string {
+  return option.description === undefined
+    ? option.label
+    : `${option.label}: ${option.description}`
+}
+
+function requireStringContent(
+  content: Record<string, ElicitationContentValue> | null | undefined,
+  key: string,
+): string | undefined {
+  const value = content?.[key]
+  return typeof value === 'string' && value.trim().length > 0 ? value : undefined
+}
+
+function askAbortError(): UserInteractionError {
+  return new UserInteractionError('ask_user_question was aborted before the user answered', 'ASK_ABORTED')
+}
+
+function withAbort<T>(promise: Promise<T>, signal: AbortSignal | undefined): Promise<T> {
+  if (signal === undefined) return promise
+  if (signal.aborted) return Promise.reject(askAbortError())
+  return new Promise<T>((resolve, reject) => {
+    const onAbort = (): void => {
+      signal.removeEventListener('abort', onAbort)
+      reject(askAbortError())
+    }
+    signal.addEventListener('abort', onAbort, { once: true })
+    promise.then(
+      (value) => {
+        signal.removeEventListener('abort', onAbort)
+        resolve(value)
+      },
+      (error: unknown) => {
+        signal.removeEventListener('abort', onAbort)
+        reject(new Error(String(error), { cause: error }))
+      },
+    )
+  })
+}
+
+function elicitationForQuestion(
+  sessionId: SessionId,
+  question: AskUserQuestionItem,
+  options: AskUserQuestionOption[],
+): CreateElicitationRequest {
+  const title = question.header ?? 'Question'
+  if (options.length === 0) {
+    return {
+      sessionId,
+      mode: 'form',
+      message: question.question,
+      requestedSchema: {
+        type: 'object',
+        title,
+        properties: {
+          custom: { type: 'string', title: question.question },
+        },
+        required: ['custom'],
+      },
+    }
+  }
+
+  const choiceOptions: EnumOption[] = options.map(option => ({
+    const: option.label,
+    title: optionDescription(option),
+  }))
+  const choice = question.multiSelect === true
+    ? {
+      type: 'array' as const,
+      title: question.question,
+      description: 'Choose one or more options, or fill a custom answer below.',
+      items: {
+        anyOf: choiceOptions,
+      },
+    }
+    : {
+      type: 'string' as const,
+      title: question.question,
+      description: 'Choose one option, or fill a custom answer below.',
+      oneOf: choiceOptions,
+    }
+  return {
+    sessionId,
+    mode: 'form',
+    message: question.question,
+    requestedSchema: {
+      type: 'object',
+      title,
+      properties: {
+        choice,
+        custom: {
+          type: 'string',
+          title: 'Custom answer',
+          description: 'Optional free-form answer. Leave empty to use the selected option.',
+        },
+      },
+      required: [],
+    },
+  }
+}
+
+function stringArrayContent(
+  content: Record<string, ElicitationContentValue> | null | undefined,
+  key: string,
+): string[] {
+  const value = content?.[key]
+  if (Array.isArray(value)) return value.filter((item): item is string => typeof item === 'string' && item.length > 0)
+  return typeof value === 'string' && value.length > 0 ? [value] : []
+}
+
 /** Plugin config: the agent template ACP sessions are created from. */
 export interface AcpConfig {
   /** Model name for created agents (must have a registered adapter). */
@@ -211,6 +332,7 @@ export function apply(ctx: Context, config: AcpConfig): void {
   const sessionPersistence = ctx.sessionPersistence
   const logger = ctx.logger
   const tools = ctx.tools
+  const userInteraction = ctx.userInteraction
   // A new ToolPresenter per session (and a throwaway per load replay), each given
   // this warn sink so a throwing tool presenter is logged, not propagated.
   const makePresenter = (): ToolPresenter => new ToolPresenter(tools, (message) => { logger.warn(message) })
@@ -241,6 +363,42 @@ export function apply(ctx: Context, config: AcpConfig): void {
   // `notify` never observes it unset — no undefined guard needed.
   let conn: AgentSideConnection
 
+  userInteraction.registerProvider({
+    async ask(request: AskUserQuestionRequest): Promise<AskUserQuestionAnswer> {
+      if (request.agent === undefined) {
+        throw new UserInteractionError('ACP user questions must come from an agent-owned request', 'NO_AGENT')
+      }
+      const sessionId = bySession.get(request.agent)
+      if (sessionId === undefined) {
+        throw new UserInteractionError('ACP user question has no matching session', 'NO_SESSION')
+      }
+      const answers: AskUserQuestionAnswerItem[] = []
+      for (const question of request.questions) {
+        const options = question.options ?? []
+        const response = await withAbort(conn.unstable_createElicitation(
+          elicitationForQuestion(sessionId, question, options),
+        ), request.signal).catch((error: unknown) => {
+          if (error instanceof UserInteractionError) throw error
+          throw new UserInteractionError('ACP elicitation request failed', 'ASK_FAILED', { cause: error })
+        })
+        if (response.action !== 'accept') {
+          throw new UserInteractionError('ask_user_question was cancelled by the user', 'ASK_CANCELLED')
+        }
+        const custom = requireStringContent(response.content, 'custom')
+        const selected = stringArrayContent(response.content, 'choice')
+        if (custom === undefined && selected.length === 0) {
+          throw new UserInteractionError('ask_user_question returned no answer', 'NO_ANSWER')
+        }
+        answers.push({
+          id: question.id,
+          selected: custom === undefined ? selected : [],
+          ...custom !== undefined ? { custom } : {},
+        })
+      }
+      return { answers }
+    },
+  })
+
   /**
    * Reject any RPC after the bridge has torn down. The `AgentSideConnection`
    * receive loop can outlive the plugin fiber — under an ACP-only HMR reload the

packages/ui/acp/tests/bridge.spec.ts

@@ -4,7 +4,7 @@ import { tmpdir } from 'node:os'
 import { join } from 'node:path'
 import { PROTOCOL_VERSION } from '@agentclientprotocol/sdk'
 import { AgentId } from '@deepseek-ai/dsh-agent'
-import { makeBridgeHarness, textResponse, type BridgeHarness } from './harness.ts'
+import { makeBridgeHarness, textResponse, toolCallResponse, type BridgeHarness } from './harness.ts'
 
 /**
  * End-to-end bridge specs over an in-memory transport: a real
@@ -53,6 +53,209 @@ describe('acp bridge', () => {
     expect(text).toBe('hello there')
   })
 
+  it('routes ask_user_question through ACP form elicitation and continues with the selected option', async () => {
+    harness = await makeBridgeHarness({
+      storageDir,
+      withAskUser: true,
+      script: [
+        toolCallResponse('ask-1', 'ask_user_question', {
+          questions: [{
+            id: 'language',
+            header: 'Project config',
+            question: 'Which language should I use?',
+            options: [
+              { label: 'TypeScript', description: 'Good for UI apps' },
+              { label: 'Python', description: 'Good for scripts' },
+            ],
+          }],
+        }),
+        textResponse('Python it is.'),
+      ],
+    })
+    harness.onElicitation = () => ({ action: 'accept', content: { choice: 'Python' } })
+    await harness.client.initialize({ protocolVersion: PROTOCOL_VERSION, clientCapabilities: {} })
+    const { sessionId } = await harness.client.newSession({ cwd: process.cwd(), mcpServers: [] })
+
+    const result = await harness.client.prompt({ sessionId, prompt: [{ type: 'text', text: 'ask me' }] })
+
+    expect(result.stopReason).toBe('end_turn')
+    expect(harness.elicitationRequests).toHaveLength(1)
+    expect(harness.elicitationRequests[0]).toMatchObject({
+      sessionId,
+      mode: 'form',
+      message: 'Which language should I use?',
+      requestedSchema: {
+        title: 'Project config',
+        properties: {
+          choice: {
+            oneOf: [
+              { const: 'TypeScript', title: 'TypeScript: Good for UI apps' },
+              { const: 'Python', title: 'Python: Good for scripts' },
+            ],
+          },
+          custom: { type: 'string' },
+        },
+        required: [],
+      },
+    })
+    const toolResult = harness.ctx.agents.get(AgentId(sessionId))!.session.events.find(event => event.type === 'tool/result')
+    const toolResultBlock = toolResult?.type === 'tool/result' ? toolResult.data.content[0] : undefined
+    const toolResultText = toolResultBlock?.type === 'text' ? toolResultBlock.text : undefined
+    expect(toolResultText).toBe('{"answers":[{"id":"language","selected":["Python"]}]}')
+  })
+
+  it('routes optionless ask_user_question through an ACP free-form answer field', async () => {
+    harness = await makeBridgeHarness({
+      storageDir,
+      withAskUser: true,
+      script: [
+        toolCallResponse('ask-1', 'ask_user_question', {
+          questions: [{ id: 'name', question: 'What should I name it?' }],
+        }),
+        textResponse('Name recorded.'),
+      ],
+    })
+    harness.onElicitation = () => ({ action: 'accept', content: { custom: 'apollo' } })
+    await harness.client.initialize({ protocolVersion: PROTOCOL_VERSION, clientCapabilities: {} })
+    const { sessionId } = await harness.client.newSession({ cwd: process.cwd(), mcpServers: [] })
+
+    await harness.client.prompt({ sessionId, prompt: [{ type: 'text', text: 'ask me' }] })
+
+    expect(harness.elicitationRequests[0]).toMatchObject({
+      requestedSchema: {
+        properties: { custom: { type: 'string', title: 'What should I name it?' } },
+        required: ['custom'],
+      },
+    })
+    const toolResult = harness.ctx.agents.get(AgentId(sessionId))!.session.events.find(event => event.type === 'tool/result')
+    expect(JSON.stringify(toolResult)).toContain('apollo')
+  })
+
+  it('supports ACP custom answers alongside choices', async () => {
+    harness = await makeBridgeHarness({ storageDir, withAskUser: true })
+    harness.onElicitation = () => ({ action: 'accept', content: { custom: 'Use Zig' } })
+    await harness.client.initialize({ protocolVersion: PROTOCOL_VERSION, clientCapabilities: {} })
+    const { sessionId } = await harness.client.newSession({ cwd: process.cwd(), mcpServers: [] })
+    const agent = harness.ctx.agents.get(AgentId(sessionId))!
+
+    const result = await harness.ctx.userInteraction.ask({
+      agent,
+      questions: [{
+        id: 'language',
+        question: 'Which language?',
+        options: [{ label: 'TypeScript' }],
+      }],
+    })
+
+    expect(result).toEqual({ answers: [{ id: 'language', selected: [], custom: 'Use Zig' }] })
+    expect(harness.elicitationRequests[0]).toMatchObject({
+      requestedSchema: {
+        properties: {
+          choice: {
+            description: 'Choose one option, or fill a custom answer below.',
+            oneOf: [{ const: 'TypeScript', title: 'TypeScript' }],
+          },
+          custom: { type: 'string' },
+        },
+        required: [],
+      },
+    })
+  })
+
+  it('treats ACP custom answers as overriding selected choices', async () => {
+    harness = await makeBridgeHarness({ storageDir, withAskUser: true })
+    harness.onElicitation = () => ({ action: 'accept', content: { choice: 'TypeScript', custom: 'Use Zig' } })
+    await harness.client.initialize({ protocolVersion: PROTOCOL_VERSION, clientCapabilities: {} })
+    const { sessionId } = await harness.client.newSession({ cwd: process.cwd(), mcpServers: [] })
+    const agent = harness.ctx.agents.get(AgentId(sessionId))!
+
+    await expect(harness.ctx.userInteraction.ask({
+      agent,
+      questions: [{
+        id: 'language',
+        question: 'Which language?',
+        options: [{ label: 'TypeScript' }],
+      }],
+    })).resolves.toEqual({ answers: [{ id: 'language', selected: [], custom: 'Use Zig' }] })
+  })
+
+  it('supports ACP multi-select answers', async () => {
+    harness = await makeBridgeHarness({ storageDir, withAskUser: true })
+    harness.onElicitation = () => ({ action: 'accept', content: { choice: ['Tests', 'Docs'] } })
+    await harness.client.initialize({ protocolVersion: PROTOCOL_VERSION, clientCapabilities: {} })
+    const { sessionId } = await harness.client.newSession({ cwd: process.cwd(), mcpServers: [] })
+    const agent = harness.ctx.agents.get(AgentId(sessionId))!
+
+    await expect(harness.ctx.userInteraction.ask({
+      agent,
+      questions: [{
+        id: 'targets',
+        question: 'Pick',
+        options: [{ label: 'Tests' }, { label: 'Docs' }],
+        multiSelect: true,
+      }],
+    })).resolves.toEqual({ answers: [{ id: 'targets', selected: ['Tests', 'Docs'] }] })
+  })
+
+  it('reports ACP ask-user routing and answer failures as structured errors', async () => {
+    harness = await makeBridgeHarness({ storageDir, withAskUser: true })
+    await harness.client.initialize({ protocolVersion: PROTOCOL_VERSION, clientCapabilities: {} })
+    const { sessionId } = await harness.client.newSession({ cwd: process.cwd(), mcpServers: [] })
+    const agent = harness.ctx.agents.get(AgentId(sessionId))!
+
+    await expect(harness.ctx.userInteraction.ask({ questions: [{ id: 'x', question: 'No agent?' }] }))
+      .rejects.toMatchObject({ name: 'UserInteractionError', code: 'NO_AGENT' })
+    await expect(harness.ctx.userInteraction.ask({ agent: { id: 'other' } as typeof agent, questions: [{ id: 'x', question: 'No session?' }] }))
+      .rejects.toMatchObject({ code: 'NO_SESSION' })
+
+    harness.onElicitation = () => ({ action: 'cancel' })
+    await expect(harness.ctx.userInteraction.ask({ agent, questions: [{ id: 'x', question: 'Cancel?' }] }))
+      .rejects.toMatchObject({ code: 'ASK_CANCELLED' })
+
+    harness.onElicitation = () => ({ action: 'accept', content: {} })
+    await expect(harness.ctx.userInteraction.ask({ agent, questions: [{ id: 'x', question: 'Empty?' }] }))
+      .rejects.toMatchObject({ code: 'NO_ANSWER' })
+
+    harness.onElicitation = () => { throw new Error('client boom') }
+    await expect(harness.ctx.userInteraction.ask({ agent, questions: [{ id: 'x', question: 'Client fails?' }], signal: new AbortController().signal }))
+      .rejects.toMatchObject({ code: 'ASK_FAILED' })
+  })
+
+  it('aborts ACP ask-user requests before and while waiting for elicitation', async () => {
+    harness = await makeBridgeHarness({ storageDir, withAskUser: true })
+    await harness.client.initialize({ protocolVersion: PROTOCOL_VERSION, clientCapabilities: {} })
+    const { sessionId } = await harness.client.newSession({ cwd: process.cwd(), mcpServers: [] })
+    const agent = harness.ctx.agents.get(AgentId(sessionId))!
+
+    const alreadyAborted = new AbortController()
+    alreadyAborted.abort()
+    await expect(harness.ctx.userInteraction.ask({ agent, questions: [{ id: 'x', question: 'Already?' }], signal: alreadyAborted.signal }))
+      .rejects.toMatchObject({ code: 'ASK_ABORTED' })
+
+    let abortedReads = 0
+    const racingAbort = {
+      get aborted() { return abortedReads++ > 0 },
+      addEventListener() {},
+      removeEventListener() {},
+      dispatchEvent() { return false },
+      onabort: null,
+      reason: undefined,
+      throwIfAborted() {},
+    } as AbortSignal
+    await expect(harness.ctx.userInteraction.ask({ agent, questions: [{ id: 'x', question: 'Raced?' }], signal: racingAbort }))
+      .rejects.toMatchObject({ code: 'ASK_ABORTED' })
+
+    let release: ((value: { action: 'accept'; content: { custom: string } }) => void) | undefined
+    harness.onElicitation = () => new Promise((resolve) => { release = resolve })
+    const pendingAbort = new AbortController()
+    const ask = harness.ctx.userInteraction.ask({ agent, questions: [{ id: 'x', question: 'Pending?' }], signal: pendingAbort.signal })
+    await new Promise(resolve => setImmediate(resolve))
+    pendingAbort.abort()
+
+    await expect(ask).rejects.toMatchObject({ code: 'ASK_ABORTED' })
+    release?.({ action: 'accept', content: { custom: 'too late' } })
+  })
+
   it('allows multiple concurrent sessions, each with a distinct id', async () => {
     harness = await makeBridgeHarness({ storageDir, script: [] })
     await harness.client.initialize({ protocolVersion: PROTOCOL_VERSION, clientCapabilities: {} })

packages/ui/acp/tests/harness.ts

@@ -29,11 +29,15 @@ import {
   ndJsonStream,
   type Agent as AcpAgent,
   type Client,
+  type CreateElicitationRequest,
+  type CreateElicitationResponse,
   type RequestPermissionRequest,
   type RequestPermissionResponse,
   type SessionNotification,
   type Stream,
 } from '@agentclientprotocol/sdk'
+import UserInteractionService from '@deepseek-ai/dsh-user-interaction'
+import * as ToolAskUser from '@deepseek-ai/dsh-tool-ask-user'
 import * as AcpPlugin from '../src/index.ts'
 import { type AcpConfig } from '../src/index.ts'
 
@@ -121,6 +125,10 @@ export interface BridgeHarness {
   permissionRequests: RequestPermissionRequest[]
   /** Decide each permission request's outcome (default: cancelled). */
   onPermission: (req: RequestPermissionRequest) => RequestPermissionResponse
+  /** Elicitation requests the bridge issued for ask_user_question. */
+  elicitationRequests: CreateElicitationRequest[]
+  /** Decide each elicitation response (default: cancel). */
+  onElicitation: (req: CreateElicitationRequest) => CreateElicitationResponse | Promise<CreateElicitationResponse>
   /** If set, the client's sessionUpdate throws this (tests notify error path). */
   onSessionUpdateError: (() => void) | undefined
   /**
@@ -164,6 +172,8 @@ export async function makeBridgeHarness(options: {
    * implementation over a mock in tests").
    */
   withBash?: boolean
+  /** Plug the REAL `ask_user_question` tool and ACP user-interaction provider. */
+  withAskUser?: boolean
   /**
    * Plug the REAL `dsh-tool-todo` tool so a test can drive `todo_write` through
    * the bridge and assert the resulting `plan` sessionUpdate — the shipping
@@ -190,6 +200,10 @@ export async function makeBridgeHarness(options: {
   await ctx.plugin(AgentRegistry)
   await ctx.plugin(AgentLoop, { agents: [] })
   await ctx.plugin(SessionPersistenceJsonl, { root: options.storageDir })
+  await ctx.plugin(UserInteractionService)
+  if (options.withAskUser) {
+    await ctx.plugin(ToolAskUser)
+  }
   if (options.withBash) {
     await ctx.plugin(LocalBashExecutor, { timeoutMs: 10_000 })
     await ctx.plugin(ToolBash)
@@ -226,6 +240,7 @@ export async function makeBridgeHarness(options: {
   const updates: CapturedUpdate[] = []
   const sessionUpdates: { sessionId: string; update: CapturedUpdate }[] = []
   const permissionRequests: RequestPermissionRequest[] = []
+  const elicitationRequests: CreateElicitationRequest[] = []
   const harness: BridgeHarness = {
     ctx,
     adapter,
@@ -233,6 +248,8 @@ export async function makeBridgeHarness(options: {
     sessionUpdates,
     permissionRequests,
     onPermission: () => ({ outcome: { outcome: 'cancelled' } }),
+    elicitationRequests,
+    onElicitation: () => ({ action: 'cancel' }),
     onSessionUpdateError: undefined,
     client: undefined as unknown as ClientSideConnection,
     acpFiber: undefined as unknown as BridgeHarness['acpFiber'],
@@ -258,6 +275,10 @@ export async function makeBridgeHarness(options: {
       permissionRequests.push(params)
       return Promise.resolve(harness.onPermission(params))
     },
+    unstable_createElicitation(params: CreateElicitationRequest): Promise<CreateElicitationResponse> {
+      elicitationRequests.push(params)
+      return Promise.resolve(harness.onElicitation(params))
+    },
   })
 
   // Wire the bridge (agent side) and the client (test side). The test config

packages/ui/acp/tsconfig.json

@@ -29,6 +29,9 @@
     {
       "path": "../../core/tools"
     },
+    {
+      "path": "../user-interaction"
+    },
     {
       "path": "../../session-persistence/session-persistence"
     }

packages/ui/stdio-agent/README.md

@@ -13,6 +13,8 @@ A terminal chat always wants the same cluster, so the package owns it rather tha
 | `@cordisjs/plugin-logger-console` | the console logger — stdout is just the terminal here, so logging to it is correct (the ACP app must NOT have this) |
 | `@deepseek-ai/dsh-agent-core` | the spine, pre-creating a `main` agent from this app's `model` and carrying its `persona` |
 | `@deepseek-ai/dsh-session-persistence-jsonl` | durable JSONL session log under `persistenceRoot` |
+| `@deepseek-ai/dsh-user-interaction` | the human question/answer seam used by confirmation tools |
+| `@deepseek-ai/dsh-tool-ask-user` | the model-facing `ask_user_question` tool |
 | `stdio-chat` (in-package module) | the readline UI, bound to the `main` agent |
 
 `@cordisjs/plugin-hmr` (the dev/demo edit-reload loop) is deliberately a **leaf** entry, NOT baked in here: it is a Loader-only, subprocess-only dev plugin — its constructor throws without `node --expose-internals` + a live `loader`, and the in-process test tier cannot even import it (so a package whose `apply` statically pulled it in could never carry the per-file coverage gate). Unlike the console logger, a stray `hmr` is not a stdout-purity footgun, so leaving it at the leaf costs no safety. The `demo:echo` / `demo:repl` leaves load it and pass `--expose-internals`.

packages/ui/stdio-agent/package.json

@@ -39,6 +39,8 @@
     "@deepseek-ai/dsh-agent-core": "^0.0.1",
     "@deepseek-ai/dsh-session": "^0.0.1",
     "@deepseek-ai/dsh-session-persistence-jsonl": "^0.0.1",
+    "@deepseek-ai/dsh-tool-ask-user": "^0.0.1",
+    "@deepseek-ai/dsh-user-interaction": "^0.0.1",
     "cordis": "^4.0.0-rc.6",
     "schemastery": "^3.17.0"
   },
@@ -53,6 +55,8 @@
     "@deepseek-ai/dsh-system-prompt": "workspace:^",
     "@deepseek-ai/dsh-session": "workspace:^",
     "@deepseek-ai/dsh-session-persistence-jsonl": "workspace:^",
+    "@deepseek-ai/dsh-tool-ask-user": "workspace:^",
+    "@deepseek-ai/dsh-user-interaction": "workspace:^",
     "cordis": "^4.0.0-rc.6",
     "schemastery": "^3.17.0"
   }

packages/ui/stdio-agent/src/index.ts

@@ -45,6 +45,8 @@ import { AgentId } from '@deepseek-ai/dsh-agent'
 import { SessionId } from '@deepseek-ai/dsh-session'
 import * as agentCore from '@deepseek-ai/dsh-agent-core'
 import SessionPersistenceJsonl from '@deepseek-ai/dsh-session-persistence-jsonl'
+import UserInteractionService from '@deepseek-ai/dsh-user-interaction'
+import * as toolAskUser from '@deepseek-ai/dsh-tool-ask-user'
 import * as uiStdio from './stdio-chat.ts'
 
 export const name = 'stdio-agent'
@@ -107,5 +109,7 @@ export function apply(ctx: Context, config: Config): void {
     }],
   })
   ctx.plugin(SessionPersistenceJsonl, { root: config.persistenceRoot ?? './.sessions' })
+  ctx.plugin(UserInteractionService)
+  ctx.plugin(toolAskUser)
   ctx.plugin(uiStdio, { welcome: config.welcome ?? 'ready.', agent: 'main' })
 }

packages/ui/stdio-agent/src/stdio-chat.ts

@@ -20,9 +20,17 @@ import type { Readable, Writable } from 'node:stream'
 import type { Context } from 'cordis'
 import z from 'schemastery'
 import { AgentId } from '@deepseek-ai/dsh-agent'
+import {
+  UserInteractionError,
+  type AskUserQuestionAnswer,
+  type AskUserQuestionAnswerItem,
+  type AskUserQuestionItem,
+  type AskUserQuestionOption,
+  type AskUserQuestionRequest,
+} from '@deepseek-ai/dsh-user-interaction'
 
 export const name = 'ui-stdio'
-export const inject = ['agents']
+export const inject = ['agents', 'userInteraction']
 
 /** Serializable plugin configuration (cordis-native, schemastery). */
 export interface Config {
@@ -57,6 +65,20 @@ function isTTYPair(input: Readable, output: Writable): boolean {
   return Boolean((input as { isTTY?: boolean }).isTTY && (output as { isTTY?: boolean }).isTTY)
 }
 
+interface PendingQuestion {
+  request: AskUserQuestionRequest
+  questionIndex: number
+  answers: AskUserQuestionAnswerItem[]
+  resolve(answer: AskUserQuestionAnswer): void
+  reject(error: unknown): void
+  onAbort: () => void
+}
+
+type OptionSelection =
+  | { kind: 'selected'; options: AskUserQuestionOption[] }
+  | { kind: 'custom' }
+  | { kind: 'invalid' }
+
 /**
  * The plugin body, parameterized over its I/O runtime. `apply` is the thin
  * production wrapper that binds the real `process` streams; tests call this
@@ -154,6 +176,8 @@ export function createStdioChat(ctx: Context, config: Config, runtime: StdioRunt
     let submittedWork = false
     let sawRunning = false
     let exitTimer: ReturnType<typeof setTimeout> | undefined
+    let activeQuestion: PendingQuestion | undefined
+    const questionQueue: PendingQuestion[] = []
 
     const maybeExit = (): void => {
       if (disposed || !stdinClosed) return
@@ -180,7 +204,152 @@ export function createStdioChat(ctx: Context, config: Config, runtime: StdioRunt
       if (status === 'idle') maybeExit()
     })
 
+    const activeQuestionItem = (pending: PendingQuestion): AskUserQuestionItem =>
+      pending.request.questions[pending.questionIndex] as AskUserQuestionItem
+
+    const renderQuestion = (pending: PendingQuestion): void => {
+      const question = activeQuestionItem(pending)
+      const options = question.options ?? []
+      output.write('\n')
+      output.write(question.header ? `[${question.header}] ${question.question}\n` : `${question.question}\n`)
+      options.forEach((option, index) => {
+        output.write(`  ${index + 1}. ${option.label}\n`)
+        if (option.description) output.write(`     ${option.description}\n`)
+      })
+      output.write('> ')
+    }
+
+    const removeAbortListener = (pending: PendingQuestion): void => {
+      pending.request.signal?.removeEventListener('abort', pending.onAbort)
+    }
+
+    const startNextQuestion = (): void => {
+      if (activeQuestion !== undefined) return
+      const pending = questionQueue.shift()
+      if (pending === undefined) return
+      // The queue never contains an aborted pending ask: the seam rejects an
+      // already-aborted request synchronously, and queued asks attach their
+      // abort listener before enqueueing.
+      activeQuestion = pending
+      renderQuestion(pending)
+    }
+
+    const disposeQuestion = (pending: PendingQuestion): void => {
+      removeAbortListener(pending)
+      pending.reject(new UserInteractionError('ask_user_question was interrupted before the user answered', 'ASK_ABORTED'))
+    }
+
+    const disposePendingQuestions = (): void => {
+      if (activeQuestion !== undefined) {
+        disposeQuestion(activeQuestion)
+        activeQuestion = undefined
+      }
+      for (const pending of questionQueue.splice(0)) {
+        disposeQuestion(pending)
+      }
+    }
+
+    const finishQuestion = (pending: PendingQuestion): void => {
+      activeQuestion = undefined
+      removeAbortListener(pending)
+      pending.resolve({ answers: pending.answers })
+      output.write('\n')
+      startNextQuestion()
+    }
+
+    const answerCurrentQuestion = (pending: PendingQuestion, answer: AskUserQuestionAnswerItem): void => {
+      pending.answers.push(answer)
+      pending.questionIndex += 1
+      if (pending.questionIndex >= pending.request.questions.length) {
+        finishQuestion(pending)
+        return
+      }
+      renderQuestion(pending)
+    }
+
+    const selectedOptions = (text: string, options: AskUserQuestionOption[], multiSelect: boolean): OptionSelection => {
+      if (text === '') return { kind: 'invalid' }
+      if (!multiSelect) {
+        if (!/^\d+$/.test(text)) return { kind: 'custom' }
+        const selected = options[Number(text) - 1]
+        return selected === undefined ? { kind: 'invalid' } : { kind: 'selected', options: [selected] }
+      }
+      const indices = text.split(/[,\s]+/).filter(Boolean)
+      if (indices.length === 0) return { kind: 'invalid' }
+      if (indices.some(part => !/^\d+$/.test(part))) return { kind: 'custom' }
+      const uniqueIndices = [...new Set(indices)]
+      const selected = uniqueIndices.map(part => options[Number(part) - 1])
+      return selected.some(option => option === undefined)
+        ? { kind: 'invalid' }
+        : { kind: 'selected', options: selected as AskUserQuestionOption[] }
+    }
+
+    const answerQuestion = (line: string): void => {
+      const pending = activeQuestion as PendingQuestion
+      const question = activeQuestionItem(pending)
+
+      const text = line.trim()
+      const options = question.options ?? []
+      const selection = options.length > 0
+        ? selectedOptions(text, options, question.multiSelect ?? false)
+        : { kind: text === '' ? 'invalid' : 'custom' } as OptionSelection
+      if (selection.kind === 'selected') {
+        answerCurrentQuestion(pending, { id: question.id, selected: selection.options.map(option => option.label) })
+        return
+      }
+
+      if (selection.kind === 'custom' && text !== '') {
+        answerCurrentQuestion(pending, { id: question.id, selected: [], custom: text })
+        return
+      }
+
+      output.write(options.length > 0
+        ? 'Please enter one of the option numbers'
+          + (question.multiSelect ? ' (comma or space separated)' : '')
+          + ' or a custom answer'
+          + '.\n> '
+        : 'Please enter an answer.\n> ')
+    }
+
+    const disposeUserInteractionProvider = ctx.userInteraction.registerProvider({
+      ask(request) {
+        if (disposed || stdinClosed) {
+          return Promise.reject(
+            new UserInteractionError('ask_user_question cannot be answered because stdin is closed', 'ASK_ABORTED'),
+          )
+        }
+        return new Promise<AskUserQuestionAnswer>((resolve, reject) => {
+          const pending: PendingQuestion = {
+            request,
+            questionIndex: 0,
+            answers: [],
+            resolve,
+            reject,
+            onAbort: () => {
+              if (activeQuestion === pending) {
+                activeQuestion = undefined
+                disposeQuestion(pending)
+                startNextQuestion()
+                return
+              }
+              // If it is not active, this listener can only fire while the ask
+              // remains queued; settled asks remove the listener first.
+              questionQueue.splice(questionQueue.indexOf(pending), 1)
+              disposeQuestion(pending)
+            },
+          }
+          request.signal?.addEventListener('abort', pending.onAbort, { once: true })
+          questionQueue.push(pending)
+          startNextQuestion()
+        })
+      },
+    })
+
     reader.on('line', (line) => {
+      if (activeQuestion !== undefined) {
+        answerQuestion(line)
+        return
+      }
       const text = line.trim()
       if (!text) return
       const agent = ctx.agents.get(agentId)
@@ -199,12 +368,15 @@ export function createStdioChat(ctx: Context, config: Config, runtime: StdioRunt
       // Fires for BOTH stdin EOF and plugin disposal (reader.close() below);
       // `disposed` guards teardown so HMR/dispose never exits the process.
       stdinClosed = true
+      if (!disposed) disposePendingQuestions()
       maybeExit()
     })
     output.write(`${welcome}\n> `)
     return () => {
       disposed = true
       if (exitTimer !== undefined) clearTimeout(exitTimer)
+      disposePendingQuestions()
+      disposeUserInteractionProvider()
       disposeStatusListener()
       reader.close()
     }

packages/ui/stdio-agent/tests/readline.spec.ts

@@ -19,6 +19,7 @@ function fakeContext(): Context {
     // The UI seeds its label map from the registry at install; this suite only
     // exercises readline terminal-mode selection, so an empty roster suffices.
     agents: { list: vi.fn(() => []) },
+    userInteraction: { registerProvider: vi.fn(() => vi.fn()) },
   } as unknown as Context
 }
 

packages/ui/stdio-agent/tests/stdio-agent.spec.ts

@@ -37,6 +37,8 @@ describe('dsh-stdio-agent app', () => {
     expect(ctx.get('agents')).toBeDefined()
     expect(ctx.get('agentLoop')).toBeDefined()
     expect(ctx.get('sessionPersistence')).toBeDefined()
+    expect(ctx.get('userInteraction')).toBeDefined()
+    expect(ctx.get('tools')?.get('ask_user_question')).toBeDefined()
     // The pre-created `main` agent the UI drives.
     expect(ctx.get('agents')?.get(AgentId('main'))).toBeDefined()
     await ctx.fiber.dispose()
@@ -92,9 +94,10 @@ describe('dsh-stdio-agent app', () => {
       })
     }
     const assembly = await ctx.get('systemPrompt')!.assemble()
-    // The rest-slot is lexicographic: the bundle's own task control tools
-    // (tool-tasks needs no executor, unlike the pending bash tool) follow alpha.
-    expect(assembly.tools.map(tool => tool.name)).toEqual(['zulu', 'alpha', 'task_kill', 'task_list', 'task_output'])
+    // The rest-slot is lexicographic: the app's ask_user_question and the
+    // bundle's own task control tools (neither needs an executor, unlike the
+    // pending bash tool) follow alpha.
+    expect(assembly.tools.map(tool => tool.name)).toEqual(['zulu', 'alpha', 'ask_user_question', 'task_kill', 'task_list', 'task_output'])
     await ctx.fiber.dispose()
   })
 

packages/ui/stdio-agent/tests/stdio-chat.spec.ts

@@ -1,10 +1,11 @@
-import { Readable } from 'node:stream'
+import { Readable, Writable } from 'node:stream'
 import { describe, expect, it, vi } from 'vitest'
 import { Context } from 'cordis'
 import type { Agent, AgentStatus } from '@deepseek-ai/dsh-agent'
 import AgentRegistry from '@deepseek-ai/dsh-agent'
 import type { ContentBlock, StreamChunk } from '@deepseek-ai/dsh-llm'
 import type { Session, SessionEvent } from '@deepseek-ai/dsh-session'
+import UserInteractionService from '@deepseek-ai/dsh-user-interaction'
 import { createStdioChat, type Config, type StdioRuntime } from '../src/stdio-chat.ts'
 
 /**
@@ -79,10 +80,11 @@ const CONFIG: Config = { welcome: 'hi there', agent: 'main' }
 async function setup(config: Config = CONFIG, runtimeOver: Partial<StdioRuntime> = {}) {
   const ctx = new Context()
   await ctx.plugin(AgentRegistry)
+  await ctx.plugin(UserInteractionService)
   const { runtime, input, out, exit } = makeRuntime(runtimeOver)
   const fiber = await ctx.plugin(Object.assign((inner: Context) => {
     createStdioChat(inner, config, runtime)
-  }, { inject: ['agents'] }))
+  }, { inject: ['agents', 'userInteraction'] }))
   return { ctx, fiber, input, out, exit }
 }
 
@@ -105,6 +107,30 @@ describe('createStdioChat rendering', () => {
     // And it drives the default agent id 'main'.
   })
 
+  it('detects readline terminal mode from both stream TTY flags', async () => {
+    for (const [inputTTY, outputTTY] of [[true, false], [true, true]] as const) {
+      const ctx = new Context()
+      await ctx.plugin(AgentRegistry)
+      await ctx.plugin(UserInteractionService)
+      let text = ''
+      const output = new Writable({
+        write(chunk, _encoding, callback) {
+          text += String(chunk)
+          callback()
+        },
+      }) as Writable & { isTTY?: boolean }
+      const { runtime } = makeRuntime({ output })
+      ;(runtime.input as Readable & { isTTY?: boolean }).isTTY = inputTTY
+      output.isTTY = outputTTY
+      const fiber = await ctx.plugin(Object.assign((inner: Context) => {
+        createStdioChat(inner, CONFIG, runtime)
+      }, { inject: ['agents', 'userInteraction'] }))
+
+      expect(text).toContain('hi there')
+      await fiber.dispose()
+    }
+  })
+
   it('renders text-delta chunks verbatim', async () => {
     const { ctx, out } = await setup()
     ctx.emit('session/event', makeSession('main'), chunkEvent({ type: 'text-delta', index: 0, text: 'hello' }))
@@ -160,12 +186,13 @@ describe('createStdioChat rendering', () => {
     // of the raw session id.
     const ctx = new Context()
     await ctx.plugin(AgentRegistry)
+    await ctx.plugin(UserInteractionService)
     const agent = makeAgent('main')
     ctx.agents.register(agent) // registered BEFORE the UI plugin below
     const { runtime, out } = makeRuntime()
     await ctx.plugin(Object.assign((inner: Context) => {
       createStdioChat(inner, CONFIG, runtime)
-    }, { inject: ['agents'] }))
+    }, { inject: ['agents', 'userInteraction'] }))
     ctx.emit('session/event', makeSession('main'), {
       type: 'turn/start', seq: 1, time: 0, data: { turn: 5, trigger: { kind: 'message' } },
     } as SessionEvent)
@@ -264,6 +291,347 @@ describe('createStdioChat rendering', () => {
 })
 
 describe('createStdioChat input', () => {
+  it('answers a pending user question instead of sending the line to the agent', async () => {
+    const { ctx, input, out } = await setup()
+    const agent = makeAgent('main', 'idle')
+    ctx.agents.register(agent)
+
+    const answer = ctx.userInteraction.ask({
+      questions: [{
+        id: 'confirm',
+        header: 'Confirm',
+        question: 'Proceed with the edit?',
+        options: [{ label: 'Yes', description: 'Apply the edit now.' }],
+      }],
+    })
+    await new Promise(r => setImmediate(r))
+    input.feed('Use a smaller change')
+
+    await expect(answer).resolves.toEqual({ answers: [{ id: 'confirm', selected: [], custom: 'Use a smaller change' }] })
+    expect(agent.sent).toEqual([])
+    expect(out.text()).toContain('[Confirm] Proceed with the edit?')
+    expect(out.text()).toContain('1. Yes')
+    expect(out.text()).toContain('Apply the edit now.')
+  })
+
+  it('answers a pending user question by numeric option selection', async () => {
+    const { ctx, input } = await setup()
+    const answer = ctx.userInteraction.ask({
+      questions: [{
+        id: 'mode',
+        question: 'Which mode?',
+        options: [
+          { label: 'Safe' },
+          { label: 'Fast' },
+        ],
+      }],
+    })
+    await new Promise(r => setImmediate(r))
+    input.feed('2')
+
+    await expect(answer).resolves.toEqual({
+      answers: [{ id: 'mode', selected: ['Fast'] }],
+    })
+  })
+
+  it('renders options in input order and selects by displayed number', async () => {
+    const { ctx, input, out } = await setup()
+    const answer = ctx.userInteraction.ask({
+      questions: [{
+        id: 'topic',
+        question: 'Which topic?',
+        options: [
+          { label: 'Hobbies' },
+          { label: 'Work', description: 'Questions about current projects.' },
+          { label: 'Casual', description: 'Easy conversation.' },
+        ],
+      }],
+    })
+    await new Promise(r => setImmediate(r))
+
+    expect(out.text()).toContain([
+      'Which topic?',
+      '  1. Hobbies',
+      '  2. Work',
+      '     Questions about current projects.',
+      '  3. Casual',
+      '     Easy conversation.',
+    ].join('\n'))
+    input.feed('3')
+
+    await expect(answer).resolves.toEqual({
+      answers: [{ id: 'topic', selected: ['Casual'] }],
+    })
+  })
+
+  it('answers a multi-select question with multiple numeric selections', async () => {
+    const { ctx, input } = await setup()
+    const answer = ctx.userInteraction.ask({
+      questions: [{
+        id: 'targets',
+        question: 'What should I update?',
+        options: [{ label: 'Tests' }, { label: 'Docs' }, { label: 'Code' }],
+        multiSelect: true,
+      }],
+    })
+    await new Promise(r => setImmediate(r))
+    input.feed('1 1, 3')
+
+    await expect(answer).resolves.toEqual({
+      answers: [{ id: 'targets', selected: ['Tests', 'Code'] }],
+    })
+  })
+
+  it('accepts non-numeric multi-select input as a custom answer', async () => {
+    const { ctx, input } = await setup()
+    const answer = ctx.userInteraction.ask({
+      questions: [{
+        id: 'targets',
+        question: 'What should I update?',
+        options: [{ label: 'Tests' }, { label: 'Docs' }],
+        multiSelect: true,
+      }],
+    })
+    await new Promise(r => setImmediate(r))
+    input.feed('the release notes')
+
+    await expect(answer).resolves.toEqual({
+      answers: [{ id: 'targets', selected: [], custom: 'the release notes' }],
+    })
+  })
+
+  it('asks every question in a batch and returns answers by id', async () => {
+    const { ctx, input, out } = await setup()
+    const answer = ctx.userInteraction.ask({
+      questions: [
+        { id: 'language', question: 'Which language?', options: [{ label: 'Python' }, { label: 'TypeScript' }] },
+        { id: 'note', question: 'Any note?' },
+      ],
+    })
+    await new Promise(r => setImmediate(r))
+    input.feed('2')
+    await new Promise(r => setImmediate(r))
+    expect(out.text()).toContain('\nAny note?\n')
+    input.feed('ship today')
+
+    await expect(answer).resolves.toEqual({
+      answers: [
+        { id: 'language', selected: ['TypeScript'] },
+        { id: 'note', selected: [], custom: 'ship today' },
+      ],
+    })
+  })
+
+  it('re-prompts when option input is invalid', async () => {
+    const { ctx, input, out } = await setup()
+    const answer = ctx.userInteraction.ask({
+      questions: [{
+        id: 'mode',
+        question: 'Which mode?',
+        options: [{ label: 'Safe' }],
+        multiSelect: true,
+      }],
+    })
+    await new Promise(r => setImmediate(r))
+    input.feed('2')
+    await new Promise(r => setImmediate(r))
+    expect(out.text()).toContain('Please enter one of the option numbers (comma or space separated) or a custom answer.')
+    input.feed('1')
+
+    await expect(answer).resolves.toEqual({
+      answers: [{ id: 'mode', selected: ['Safe'] }],
+    })
+  })
+
+  it('re-prompts when single-select option input is out of range', async () => {
+    const { ctx, input, out } = await setup()
+    const answer = ctx.userInteraction.ask({
+      questions: [{
+        id: 'mode',
+        question: 'Which mode?',
+        options: [{ label: 'Safe' }],
+      }],
+    })
+    await new Promise(r => setImmediate(r))
+    input.feed('2')
+    await new Promise(r => setImmediate(r))
+    expect(out.text()).toContain('Please enter one of the option numbers or a custom answer.')
+    input.feed('1')
+
+    await expect(answer).resolves.toEqual({
+      answers: [{ id: 'mode', selected: ['Safe'] }],
+    })
+  })
+
+  it('re-prompts when multi-select input contains no option numbers', async () => {
+    const { ctx, input, out } = await setup()
+    const answer = ctx.userInteraction.ask({
+      questions: [{
+        id: 'mode',
+        question: 'Which mode?',
+        options: [{ label: 'Safe' }],
+        multiSelect: true,
+      }],
+    })
+    await new Promise(r => setImmediate(r))
+    input.feed(',')
+    await new Promise(r => setImmediate(r))
+    expect(out.text()).toContain('Please enter one of the option numbers (comma or space separated) or a custom answer.')
+    input.feed('1')
+
+    await expect(answer).resolves.toEqual({
+      answers: [{ id: 'mode', selected: ['Safe'] }],
+    })
+  })
+
+  it('re-prompts when an option question receives an empty answer', async () => {
+    const { ctx, input, out } = await setup()
+    const answer = ctx.userInteraction.ask({
+      questions: [{
+        id: 'mode',
+        question: 'Which mode?',
+        options: [{ label: 'Safe' }],
+      }],
+    })
+    await new Promise(r => setImmediate(r))
+    input.feed('')
+    await new Promise(r => setImmediate(r))
+    expect(out.text()).toContain('Please enter one of the option numbers or a custom answer.')
+    input.feed('1')
+
+    await expect(answer).resolves.toEqual({
+      answers: [{ id: 'mode', selected: ['Safe'] }],
+    })
+  })
+
+  it('re-prompts when a question receives an empty answer', async () => {
+    const { ctx, input, out } = await setup()
+    const answer = ctx.userInteraction.ask({ questions: [{ id: 'path', question: 'What should I use?' }] })
+    await new Promise(r => setImmediate(r))
+    input.feed('')
+    await new Promise(r => setImmediate(r))
+    expect(out.text()).toContain('Please enter an answer.')
+    input.feed('Use defaults')
+
+    await expect(answer).resolves.toEqual({ answers: [{ id: 'path', selected: [], custom: 'Use defaults' }] })
+  })
+
+  it('rejects an active question when its signal aborts', async () => {
+    const { ctx } = await setup()
+    const controller = new AbortController()
+    const answer = ctx.userInteraction.ask({ questions: [{ id: 'continue', question: 'Continue?' }], signal: controller.signal })
+    const rejected = expect(answer).rejects.toMatchObject({ code: 'ASK_ABORTED' })
+    await new Promise(r => setImmediate(r))
+
+    controller.abort()
+
+    await rejected
+  })
+
+  it('continues to the next queued question when the active question aborts', async () => {
+    const { ctx, input, out } = await setup()
+    const controller = new AbortController()
+    const first = ctx.userInteraction.ask({ questions: [{ id: 'first', question: 'First?' }], signal: controller.signal })
+    const firstRejected = expect(first).rejects.toMatchObject({ code: 'ASK_ABORTED' })
+    const second = ctx.userInteraction.ask({ questions: [{ id: 'second', question: 'Second?' }] })
+    await new Promise(r => setImmediate(r))
+
+    controller.abort()
+    await firstRejected
+    await new Promise(r => setImmediate(r))
+    expect(out.text()).toContain('\nSecond?\n')
+    input.feed('second answer')
+
+    await expect(second).resolves.toEqual({ answers: [{ id: 'second', selected: [], custom: 'second answer' }] })
+  })
+
+  it('skips a queued question whose signal aborted before it became active', async () => {
+    const { ctx, input, out } = await setup()
+    const controller = new AbortController()
+    const first = ctx.userInteraction.ask({ questions: [{ id: 'first', question: 'First?' }] })
+    const second = ctx.userInteraction.ask({ questions: [{ id: 'second', question: 'Second?' }], signal: controller.signal })
+    await new Promise(r => setImmediate(r))
+
+    controller.abort()
+
+    await expect(Promise.race([
+      second.then(
+        () => 'resolved',
+        (error: unknown) => (error as { code?: string }).code,
+      ),
+      new Promise<string>((resolve) => { setImmediate(() => { resolve('pending') }) }),
+    ])).resolves.toBe('ASK_ABORTED')
+    expect(out.text()).not.toContain('\nSecond?\n')
+    input.feed('first answer')
+    await expect(first).resolves.toEqual({ answers: [{ id: 'first', selected: [], custom: 'first answer' }] })
+  })
+
+  it('removes an aborted queued question without promoting later queued work early', async () => {
+    const { ctx, input, out } = await setup()
+    const controller = new AbortController()
+    const first = ctx.userInteraction.ask({ questions: [{ id: 'first', question: 'First?' }] })
+    const second = ctx.userInteraction.ask({ questions: [{ id: 'second', question: 'Second?' }], signal: controller.signal })
+    const third = ctx.userInteraction.ask({ questions: [{ id: 'third', question: 'Third?' }] })
+    await new Promise(r => setImmediate(r))
+
+    controller.abort()
+
+    await expect(second).rejects.toMatchObject({ code: 'ASK_ABORTED' })
+    expect(out.text()).toContain('\nFirst?\n')
+    expect(out.text()).not.toContain('\nSecond?\n')
+    expect(out.text()).not.toContain('\nThird?\n')
+    input.feed('first answer')
+    await new Promise(r => setImmediate(r))
+
+    expect(out.text()).toContain('\nThird?\n')
+    input.feed('third answer')
+
+    await expect(first).resolves.toEqual({ answers: [{ id: 'first', selected: [], custom: 'first answer' }] })
+    await expect(third).resolves.toEqual({ answers: [{ id: 'third', selected: [], custom: 'third answer' }] })
+  })
+
+  it('rejects active and queued questions when the UI is disposed', async () => {
+    const { ctx, fiber } = await setup()
+    const active = ctx.userInteraction.ask({ questions: [{ id: 'active', question: 'Active?' }] })
+    const queued = ctx.userInteraction.ask({ questions: [{ id: 'queued', question: 'Queued?' }] })
+    const activeRejected = expect(active).rejects.toMatchObject({ code: 'ASK_ABORTED' })
+    const queuedRejected = expect(queued).rejects.toMatchObject({ code: 'ASK_ABORTED' })
+    await new Promise(r => setImmediate(r))
+
+    await fiber.dispose()
+
+    await activeRejected
+    await queuedRejected
+  })
+
+  it('rejects active and queued questions when stdin closes before the user answers', async () => {
+    const { ctx, input, exit } = await setup()
+    const active = ctx.userInteraction.ask({ questions: [{ id: 'active', question: 'Active?' }] })
+    const queued = ctx.userInteraction.ask({ questions: [{ id: 'queued', question: 'Queued?' }] })
+    const activeRejected = expect(active).rejects.toMatchObject({ code: 'ASK_ABORTED' })
+    const queuedRejected = expect(queued).rejects.toMatchObject({ code: 'ASK_ABORTED' })
+    await new Promise(r => setImmediate(r))
+
+    input.finish()
+    await new Promise(r => setImmediate(r))
+
+    await activeRejected
+    await queuedRejected
+    expect(exit).not.toHaveBeenCalled()
+  })
+
+  it('rejects new questions immediately after stdin has closed', async () => {
+    const { ctx, input, out } = await setup()
+    input.finish()
+    await new Promise(r => setImmediate(r))
+    const before = out.text()
+
+    const answer = ctx.userInteraction.ask({ questions: [{ id: 'late', question: 'Too late?' }] })
+
+    await expect(answer).rejects.toMatchObject({ code: 'ASK_ABORTED' })
+    expect(out.text()).toBe(before)
+  })
+
   it('sends a typed line to an idle agent', async () => {
     const { ctx, input } = await setup()
     const agent = makeAgent('main', 'idle')

packages/ui/stdio-agent/tsconfig.json

@@ -32,6 +32,12 @@
     {
       "path": "../../core/agent-core"
     },
+    {
+      "path": "../user-interaction"
+    },
+    {
+      "path": "../tool-ask-user"
+    },
     {
       "path": "../../session-persistence/session-persistence-jsonl"
     }

packages/ui/tool-ask-user/README.md

@@ -0,0 +1,20 @@
+# @deepseek-ai/dsh-tool-ask-user
+
+Model-facing `ask_user_question` tool over `ctx.userInteraction`. It lets the model ask the human a concise question when it needs confirmation, a choice, or missing information before continuing.
+
+## Tool
+
+`ask_user_question` accepts:
+
+- `questions` — required non-empty array of question objects.
+- `id` — required stable id on each question, echoed in the answer.
+- `question` — required question text for each question.
+- `header` — optional short heading.
+- `options` — optional choices with `label` and `description`. If recommending a choice, put it first and append `(Recommended)` to that label.
+- `multi_select` — whether that question may return more than one selected option.
+
+The tool calls `ctx.userInteraction.ask()` and returns JSON text shaped as `{ "answers": [{ "id": "...", "selected": ["..."], "custom": "..." }] }`. `selected` contains option labels; `custom` is present only for a free-form answer and overrides selected choices.
+
+## Role
+
+This is the consumer package for the user-interaction seam. It does not render UI and does not know how input is collected; it only translates model arguments into `AskUserQuestionRequest` and returns the human answer to the agent loop.

packages/ui/tool-ask-user/package.json

@@ -0,0 +1,38 @@
+{
+  "name": "@deepseek-ai/dsh-tool-ask-user",
+  "description": "Model-facing ask_user_question tool over the ctx.userInteraction seam",
+  "version": "0.0.1",
+  "private": true,
+  "type": "module",
+  "main": "lib/index.js",
+  "types": "lib/types/index.d.ts",
+  "exports": {
+    ".": {
+      "types": "./lib/types/index.d.ts",
+      "default": "./lib/index.js"
+    },
+    "./src/*": "./src/*",
+    "./package.json": "./package.json"
+  },
+  "files": [
+    "lib/index.js",
+    "lib/types/**/*.d.ts",
+    "lib/types/**/*.d.ts.map",
+    "src"
+  ],
+  "license": "BSD-3-Clause",
+  "peerDependencies": {
+    "@deepseek-ai/dsh-agent": "^0.0.1",
+    "@deepseek-ai/dsh-tools": "^0.0.1",
+    "@deepseek-ai/dsh-user-interaction": "^0.0.1",
+    "cordis": "^4.0.0-rc.6"
+  },
+  "devDependencies": {
+    "@deepseek-ai/dsh-agent": "workspace:^",
+    "@deepseek-ai/dsh-llm": "workspace:^",
+    "@deepseek-ai/dsh-system-prompt": "workspace:^",
+    "@deepseek-ai/dsh-tools": "workspace:^",
+    "@deepseek-ai/dsh-user-interaction": "workspace:^",
+    "cordis": "^4.0.0-rc.6"
+  }
+}

packages/ui/tool-ask-user/src/index.ts

@@ -0,0 +1,71 @@
+/**
+ * Model-facing `ask_user_question` tool over the `ctx.userInteraction` seam.
+ * The tool pauses until a UI provider returns a human answer, then feeds that
+ * answer back into the agent loop as an ordinary tool result.
+ *
+ * @module @deepseek-ai/dsh-tool-ask-user
+ */
+
+import type { Context } from 'cordis'
+import { defineTool } from '@deepseek-ai/dsh-tools'
+import '@deepseek-ai/dsh-user-interaction'
+
+export const name = 'tool-ask-user'
+export const inject = ['tools', 'userInteraction']
+
+const description = 'Ask the user a concise question when you need confirmation, a choice, or missing information before proceeding. '
+  + 'Send one or more questions, each with a stable id that will be echoed in the answer.'
+
+export function apply(ctx: Context): void {
+  ctx.tools.register(defineTool({
+    name: 'ask_user_question',
+    description,
+    parameters: {
+      questions: {
+        type: 'array',
+        required: true,
+        description: 'Questions to ask the user before continuing.',
+        items: {
+          type: 'object',
+          properties: {
+            id: { type: 'string', required: true, description: 'Stable id for this question; echoed in the answer.' },
+            question: { type: 'string', required: true, description: 'The specific question to ask the user.' },
+            header: {
+              type: 'string',
+              description: 'Optional short heading for the question, such as "Confirm" or "Choose Mode".',
+            },
+            options: {
+              type: 'array',
+              description: 'Optional choices to show the user. If you recommend one, put it first and append "(Recommended)" to that label.',
+              items: {
+                type: 'object',
+                properties: {
+                  label: { type: 'string', required: true, description: 'Short user-facing option label.' },
+                  description: { type: 'string', description: 'One sentence explaining the tradeoff or impact.' },
+                },
+              },
+            },
+            multi_select: {
+              type: 'boolean',
+              description: 'Whether the user may select more than one option. Defaults to false.',
+            },
+          },
+        },
+      },
+    },
+    async execute(args, exec) {
+      const result = await ctx.userInteraction.ask({
+        questions: args.questions.map(question => ({
+          id: question.id,
+          question: question.question,
+          ...question.header !== undefined ? { header: question.header } : {},
+          ...question.options !== undefined ? { options: question.options } : {},
+          ...question.multi_select !== undefined ? { multiSelect: question.multi_select } : {},
+        })),
+        ...exec.agent !== undefined ? { agent: exec.agent } : {},
+        ...exec.signal !== undefined ? { signal: exec.signal } : {},
+      })
+      return [{ type: 'text', text: JSON.stringify(result) }]
+    },
+  }))
+}

packages/ui/tool-ask-user/tests/tool-ask-user.spec.ts

@@ -0,0 +1,253 @@
+import { describe, expect, it } from 'vitest'
+import { Context } from 'cordis'
+import { CallId } from '@deepseek-ai/dsh-llm'
+import type { Agent } from '@deepseek-ai/dsh-agent'
+import SystemPrompt from '@deepseek-ai/dsh-system-prompt'
+import ToolRegistry from '@deepseek-ai/dsh-tools'
+import UserInteractionService, { type AskUserQuestionRequest } from '@deepseek-ai/dsh-user-interaction'
+import * as toolAskUser from '@deepseek-ai/dsh-tool-ask-user'
+
+interface OptionSchemaShape {
+  properties: {
+    questions: {
+      items: {
+        properties: {
+          options: {
+            items: {
+              properties: Record<string, { type: string }>
+            }
+          }
+        } & Record<string, unknown>
+      }
+    }
+  }
+}
+
+async function setup() {
+  const ctx = new Context()
+  await ctx.plugin(SystemPrompt)
+  await ctx.plugin(ToolRegistry)
+  await ctx.plugin(UserInteractionService)
+  await ctx.plugin(toolAskUser)
+  return ctx
+}
+
+describe('ask_user_question tool', () => {
+  it('registers a model-facing tool schema', async () => {
+    const ctx = await setup()
+    const schema = ctx.tools.schemas().find(tool => tool.name === 'ask_user_question')
+
+    expect(schema).toMatchObject({
+      name: 'ask_user_question',
+      parameters: {
+        type: 'object',
+        properties: {
+          questions: { type: 'array' },
+        },
+        required: ['questions'],
+      },
+    })
+    const parameters = schema?.parameters as unknown as OptionSchemaShape
+    expect(parameters.properties.questions.items.properties).toMatchObject({
+      id: { type: 'string' },
+      question: { type: 'string' },
+      header: { type: 'string' },
+      options: { type: 'array' },
+      multi_select: { type: 'boolean' },
+    })
+    expect(parameters.properties.questions.items.properties.options.items.properties).toMatchObject({
+      label: { type: 'string' },
+      description: { type: 'string' },
+    })
+    expect(parameters.properties.questions.items.properties.options.items.properties).not.toHaveProperty('value')
+    expect(parameters.properties.questions.items.properties.options.items.properties).not.toHaveProperty('recommended')
+    expect(parameters.properties.questions.items.properties.options.items.properties).not.toHaveProperty('preview')
+  })
+
+  it('asks the registered user-interaction provider and projects structured answers to text', async () => {
+    const ctx = await setup()
+    const seen: AskUserQuestionRequest[] = []
+    ctx.userInteraction.registerProvider({
+      async ask(request) {
+        seen.push(request)
+        return { answers: [{ id: 'pkg', selected: ['pnpm'] }] }
+      },
+    })
+
+    const result = await ctx.tools.execute({
+      callId: CallId('ask-1'),
+      name: 'ask_user_question',
+      arguments: {
+        questions: [{
+          id: 'pkg',
+          question: 'Which package manager should I use?',
+          options: [{ label: 'pnpm', description: 'Use pnpm workspaces.' }],
+        }],
+      },
+    })
+
+    expect(result).toMatchObject({
+      isError: false,
+      content: [{ type: 'text', text: '{"answers":[{"id":"pkg","selected":["pnpm"]}]}' }],
+    })
+    expect(seen).toMatchObject([{
+      questions: [{
+        id: 'pkg',
+        question: 'Which package manager should I use?',
+        options: [{ label: 'pnpm', description: 'Use pnpm workspaces.' }],
+      }],
+    }])
+  })
+
+  it('passes recommended option labels through without adding schema fields', async () => {
+    const ctx = await setup()
+    const seen: AskUserQuestionRequest[] = []
+    ctx.userInteraction.registerProvider({
+      async ask(request) {
+        seen.push(request)
+        return { answers: [{ id: 'pkg', selected: ['pnpm (Recommended)'] }] }
+      },
+    })
+
+    await ctx.tools.execute({
+      callId: CallId('ask-recommended'),
+      name: 'ask_user_question',
+      arguments: {
+        questions: [{
+          id: 'pkg',
+          question: 'Which package manager should I use?',
+          options: [
+            { label: 'pnpm (Recommended)' },
+            { label: 'npm' },
+          ],
+        }],
+      },
+    })
+
+    expect(seen[0]?.questions[0]?.options).toEqual([
+      { label: 'pnpm (Recommended)' },
+      { label: 'npm' },
+    ])
+  })
+
+  it('projects custom answers and multi-select choices', async () => {
+    const ctx = await setup()
+    ctx.userInteraction.registerProvider({
+      async ask() {
+        return {
+          answers: [
+            { id: 'targets', selected: ['tests', 'docs'] },
+            { id: 'notes', selected: [], custom: 'ship today' },
+          ],
+        }
+      },
+    })
+
+    const result = await ctx.tools.execute({
+      callId: CallId('ask-multi'),
+      name: 'ask_user_question',
+      arguments: {
+        questions: [
+          {
+            id: 'targets',
+            question: 'What should I update?',
+            options: [{ label: 'tests' }, { label: 'docs' }],
+            multi_select: true,
+          },
+          { id: 'notes', question: 'Any note?' },
+        ],
+      },
+    })
+
+    expect(result.content).toEqual([{
+      type: 'text',
+      text: '{"answers":[{"id":"targets","selected":["tests","docs"]},{"id":"notes","selected":[],"custom":"ship today"}]}',
+    }])
+  })
+
+  it('passes the tool abort signal to the user-interaction request', async () => {
+    const ctx = await setup()
+    const seen: AskUserQuestionRequest[] = []
+    ctx.userInteraction.registerProvider({
+      async ask(request) {
+        seen.push(request)
+        return { answers: [{ id: 'continue', selected: ['ok'] }] }
+      },
+    })
+    const controller = new AbortController()
+
+    await ctx.tools.execute({
+      callId: CallId('ask-2'),
+      name: 'ask_user_question',
+      arguments: { questions: [{ id: 'continue', question: 'Continue?' }] },
+      signal: controller.signal,
+    })
+
+    expect(seen[0]?.signal).toBe(controller.signal)
+  })
+
+  it('passes optional header and agent through to the user-interaction request', async () => {
+    const ctx = await setup()
+    const seen: AskUserQuestionRequest[] = []
+    ctx.userInteraction.registerProvider({
+      async ask(request) {
+        seen.push(request)
+        return { answers: [{ id: 'continue', selected: ['ok'] }] }
+      },
+    })
+    const agent = { id: 'main' } as unknown as Agent
+
+    const result = await ctx.tools.execute({
+      callId: CallId('ask-3'),
+      name: 'ask_user_question',
+      arguments: { questions: [{ id: 'continue', header: 'Confirm', question: 'Continue?' }] },
+      agent,
+    })
+
+    expect(result.content).toEqual([{ type: 'text', text: '{"answers":[{"id":"continue","selected":["ok"]}]}' }])
+    expect(seen[0]).toMatchObject({ questions: [{ id: 'continue', header: 'Confirm', question: 'Continue?' }], agent })
+  })
+
+  it('returns structured user-interaction errors through tool execution', async () => {
+    const ctx = await setup()
+
+    const result = await ctx.tools.execute({
+      callId: CallId('ask-no-provider'),
+      name: 'ask_user_question',
+      arguments: { questions: [{ id: 'continue', question: 'Continue?' }] },
+    })
+
+    expect(result).toMatchObject({
+      isError: true,
+      error: { name: 'UserInteractionError', code: 'NO_PROVIDER' },
+    })
+  })
+
+  it('returns a structured error for empty question batches', async () => {
+    const ctx = await setup()
+
+    const result = await ctx.tools.execute({
+      callId: CallId('ask-empty'),
+      name: 'ask_user_question',
+      arguments: { questions: [] },
+    })
+
+    expect(result).toMatchObject({
+      isError: true,
+      error: { name: 'UserInteractionError', code: 'EMPTY_QUESTIONS' },
+    })
+  })
+
+  it('unregisters the tool when its plugin fiber is disposed', async () => {
+    const ctx = new Context()
+    await ctx.plugin(SystemPrompt)
+    await ctx.plugin(ToolRegistry)
+    await ctx.plugin(UserInteractionService)
+    const fiber = await ctx.plugin(toolAskUser)
+    expect(ctx.tools.get('ask_user_question')).toBeDefined()
+
+    await fiber.dispose()
+
+    expect(ctx.tools.get('ask_user_question')).toBeUndefined()
+  })
+})

packages/ui/tool-ask-user/tsconfig.json

@@ -0,0 +1,36 @@
+{
+  "extends": "../../../tsconfig.base.json",
+  "compilerOptions": {
+    "rootDir": "src",
+    "outDir": "lib/types"
+  },
+  "include": [
+    "src"
+  ],
+  "references": [
+    {
+      "path": "../../../vendor/cosmokit"
+    },
+    {
+      "path": "../../../vendor/cordis"
+    },
+    {
+      "path": "../../../vendor/schemastery"
+    },
+    {
+      "path": "../../llm/llm"
+    },
+    {
+      "path": "../../core/agent"
+    },
+    {
+      "path": "../../core/system-prompt"
+    },
+    {
+      "path": "../../core/tools"
+    },
+    {
+      "path": "../user-interaction"
+    }
+  ]
+}

packages/ui/user-interaction/README.md

@@ -0,0 +1,24 @@
+# @deepseek-ai/dsh-user-interaction
+
+Abstract user-interaction seam. It owns `ctx.userInteraction`, the service a model-facing tool or permission plugin uses when it needs to pause work and ask the human for a decision.
+
+## Service: `UserInteractionService` (ctx key: `userInteraction`)
+
+### Public API
+
+- `ctx.userInteraction.registerProvider(provider): () => void` Register the UI-side provider. Only one provider may be active in a context; disposal unregisters it.
+- `ctx.userInteraction.ask(request): Promise<AskUserQuestionAnswer>` Ask the active provider and wait for the answer.
+
+### Key Types
+
+- `AskUserQuestionRequest` — `{ questions: [{ id, question, header?, options?, multiSelect? }], agent?, signal? }`.
+- `AskUserQuestionOption` — `{ label, description? }`.
+- `AskUserQuestionAnswer` — `{ answers: [{ id, selected, custom? }] }`.
+- `UserInteractionProvider` — UI implementation with `ask(request)`.
+- `UserInteractionError` — `HarnessError` subclass with codes such as `EMPTY_QUESTIONS`, `NO_PROVIDER`, `DUPLICATE_PROVIDER`, and `ASK_ABORTED`.
+
+When an answer includes `custom`, `selected` is empty; custom text is an override rather than a supplement to selected choices.
+
+## Role
+
+This is the interface package. Model-facing consumers such as `@deepseek-ai/dsh-tool-ask-user` depend on this seam; UI front doors such as the `stdio-agent` readline module and the `acp` bridge provide the provider. The loop stays unchanged: a tool call simply awaits a promise, and the tool result resumes the normal agent loop.

packages/ui/user-interaction/package.json

@@ -0,0 +1,34 @@
+{
+  "name": "@deepseek-ai/dsh-user-interaction",
+  "description": "Abstract user-interaction seam (ctx.userInteraction) for asking the human during agent runs",
+  "version": "0.0.1",
+  "private": true,
+  "type": "module",
+  "main": "lib/index.js",
+  "types": "lib/types/index.d.ts",
+  "exports": {
+    ".": {
+      "types": "./lib/types/index.d.ts",
+      "default": "./lib/index.js"
+    },
+    "./src/*": "./src/*",
+    "./package.json": "./package.json"
+  },
+  "files": [
+    "lib/index.js",
+    "lib/types/**/*.d.ts",
+    "lib/types/**/*.d.ts.map",
+    "src"
+  ],
+  "license": "BSD-3-Clause",
+  "peerDependencies": {
+    "@deepseek-ai/dsh-agent": "^0.0.1",
+    "@deepseek-ai/dsh-llm": "^0.0.1",
+    "cordis": "^4.0.0-rc.6"
+  },
+  "devDependencies": {
+    "@deepseek-ai/dsh-agent": "workspace:^",
+    "@deepseek-ai/dsh-llm": "workspace:^",
+    "cordis": "^4.0.0-rc.6"
+  }
+}

packages/ui/user-interaction/src/index.ts

@@ -0,0 +1,128 @@
+/**
+ * User-interaction seam (`ctx.userInteraction`): a UI-backed service for
+ * pausing an agent tool call until the human answers a question. The model-
+ * facing tool lives in `@deepseek-ai/dsh-tool-ask-user`; UI packages provide
+ * the single active provider.
+ *
+ * @module @deepseek-ai/dsh-user-interaction
+ */
+
+import { Context, Service } from 'cordis'
+import type { Agent } from '@deepseek-ai/dsh-agent'
+import { HarnessError } from '@deepseek-ai/dsh-llm'
+
+declare module 'cordis' {
+  interface Context {
+    userInteraction: UserInteractionService
+  }
+}
+
+/** One selectable answer offered to the user. */
+export interface AskUserQuestionOption {
+  /** User-facing label. */
+  label: string
+  /** Optional extra context rendered by capable UIs. */
+  description?: string
+}
+
+/** One question in an ask_user_question request. */
+export interface AskUserQuestionItem {
+  /** Stable model-provided question id, echoed in the answer. */
+  id: string
+  /** The question to display. */
+  question: string
+  /** Optional short heading/group label. */
+  header?: string
+  /** Optional choices the UI can render as a menu. */
+  options?: AskUserQuestionOption[]
+  /** Whether more than one option may be selected. Defaults to single-select. */
+  multiSelect?: boolean
+}
+
+/** Request for a human answer. */
+export interface AskUserQuestionRequest {
+  /** Questions to display. */
+  questions: AskUserQuestionItem[]
+  /** Calling agent, when the request came from an agent tool call. */
+  agent?: Agent
+  /** Abort signal for the owning tool/step. */
+  signal?: AbortSignal
+}
+
+/** Answer to one question. */
+export interface AskUserQuestionAnswerItem {
+  /** The answered question id. */
+  id: string
+  /** Selected option labels. Empty when the answer is purely custom text. */
+  selected: string[]
+  /** Optional free-text "Other" answer. */
+  custom?: string
+}
+
+/** The human's answer. */
+export interface AskUserQuestionAnswer {
+  /** Structured answers keyed by question id. */
+  answers: AskUserQuestionAnswerItem[]
+}
+
+/** UI-side provider for user questions. */
+export interface UserInteractionProvider {
+  ask(request: AskUserQuestionRequest): Promise<AskUserQuestionAnswer>
+}
+
+/** Stable error taxonomy for user-interaction failures. */
+export class UserInteractionError extends HarnessError {
+  constructor(message: string, code: string, options?: ErrorOptions) {
+    super(message, code, options)
+    this.name = 'UserInteractionError'
+  }
+}
+
+/** `ctx.userInteraction`: one active UI provider plus an `ask()` surface. */
+export class UserInteractionService extends Service {
+  private provider: UserInteractionProvider | undefined
+
+  constructor(ctx: Context) {
+    super(ctx, 'userInteraction')
+  }
+
+  /**
+   * Register the UI provider. Only one provider may be active in a context.
+   *
+   * @param provider UI-side implementation that collects answers.
+   * @returns Disposer that unregisters this provider.
+   */
+  registerProvider(provider: UserInteractionProvider): () => void {
+    const dispose = this.ctx.effect(function* (this: UserInteractionService) {
+      if (this.provider !== undefined) {
+        throw new UserInteractionError('a user-interaction provider is already registered', 'DUPLICATE_PROVIDER')
+      }
+      this.provider = provider
+      yield () => {
+        this.provider = undefined
+      }
+    }.bind(this), 'userInteraction.registerProvider()')
+    return () => void dispose()
+  }
+
+  /**
+   * Ask the active UI provider and wait for the user's answer.
+   *
+   * @param request Questions, owner agent, and abort signal.
+   * @returns The answer chosen or typed by the human.
+   */
+  async ask(request: AskUserQuestionRequest): Promise<AskUserQuestionAnswer> {
+    if (request.signal?.aborted) {
+      throw new UserInteractionError('ask_user_question was aborted before the user answered', 'ASK_ABORTED')
+    }
+    if (request.questions.length === 0) {
+      throw new UserInteractionError('ask_user_question requires at least one question', 'EMPTY_QUESTIONS')
+    }
+    if (this.provider === undefined) {
+      throw new UserInteractionError('no user-interaction provider is registered', 'NO_PROVIDER')
+    }
+    return this.provider.ask(request)
+  }
+}
+
+export default UserInteractionService

packages/ui/user-interaction/tests/user-interaction.spec.ts

@@ -0,0 +1,86 @@
+import { describe, expect, it, vi } from 'vitest'
+import { Context } from 'cordis'
+import UserInteractionService, {
+  UserInteractionError,
+  type AskUserQuestionRequest,
+  type UserInteractionProvider,
+} from '@deepseek-ai/dsh-user-interaction'
+
+function provider(answer = 'approved'): UserInteractionProvider & { seen: AskUserQuestionRequest[] } {
+  const seen: AskUserQuestionRequest[] = []
+  return {
+    seen,
+    async ask(request) {
+      seen.push(request)
+      return { answers: [{ id: request.questions[0]?.id ?? 'missing', selected: [answer] }] }
+    },
+  }
+}
+
+describe('UserInteractionService', () => {
+  it('delegates ask requests to the registered provider', async () => {
+    const ctx = new Context()
+    await ctx.plugin(UserInteractionService)
+    const p = provider('yes')
+    ctx.userInteraction.registerProvider(p)
+
+    const result = await ctx.userInteraction.ask({ questions: [{ id: 'confirm', question: 'Proceed?' }] })
+
+    expect(result).toEqual({ answers: [{ id: 'confirm', selected: ['yes'] }] })
+    expect(p.seen).toEqual([{ questions: [{ id: 'confirm', question: 'Proceed?' }] }])
+  })
+
+  it('rejects ask requests when no provider is registered', async () => {
+    const ctx = new Context()
+    await ctx.plugin(UserInteractionService)
+
+    await expect(ctx.userInteraction.ask({ questions: [{ id: 'confirm', question: 'Proceed?' }] }))
+      .rejects.toMatchObject({ name: 'UserInteractionError', code: 'NO_PROVIDER' })
+  })
+
+  it('registers providers with HMR-safe disposal', async () => {
+    const ctx = new Context()
+    await ctx.plugin(UserInteractionService)
+    const p = provider()
+    const dispose = ctx.userInteraction.registerProvider(p)
+
+    dispose()
+    dispose()
+
+    await expect(ctx.userInteraction.ask({ questions: [{ id: 'confirm', question: 'Proceed?' }] }))
+      .rejects.toMatchObject({ code: 'NO_PROVIDER' })
+  })
+
+  it('rejects duplicate providers instead of replacing the active UI', async () => {
+    const ctx = new Context()
+    await ctx.plugin(UserInteractionService)
+    ctx.userInteraction.registerProvider(provider('first'))
+
+    expect(() => ctx.userInteraction.registerProvider(provider('second')))
+      .toThrow(UserInteractionError)
+  })
+
+  it('fails before reaching the provider when the signal is already aborted', async () => {
+    const ctx = new Context()
+    await ctx.plugin(UserInteractionService)
+    const p = { ask: vi.fn(async () => ({ answers: [{ id: 'confirm', selected: ['too late'] }] })) }
+    ctx.userInteraction.registerProvider(p)
+    const controller = new AbortController()
+    controller.abort()
+
+    await expect(ctx.userInteraction.ask({ questions: [{ id: 'confirm', question: 'Proceed?' }], signal: controller.signal }))
+      .rejects.toMatchObject({ code: 'ASK_ABORTED' })
+    expect(p.ask).not.toHaveBeenCalled()
+  })
+
+  it('rejects empty question batches before reaching the provider', async () => {
+    const ctx = new Context()
+    await ctx.plugin(UserInteractionService)
+    const p = { ask: vi.fn(async () => ({ answers: [] })) }
+    ctx.userInteraction.registerProvider(p)
+
+    await expect(ctx.userInteraction.ask({ questions: [] }))
+      .rejects.toMatchObject({ name: 'UserInteractionError', code: 'EMPTY_QUESTIONS' })
+    expect(p.ask).not.toHaveBeenCalled()
+  })
+})

packages/ui/user-interaction/tsconfig.json

@@ -0,0 +1,24 @@
+{
+  "extends": "../../../tsconfig.base.json",
+  "compilerOptions": {
+    "rootDir": "src",
+    "outDir": "lib/types"
+  },
+  "include": [
+    "src"
+  ],
+  "references": [
+    {
+      "path": "../../../vendor/cosmokit"
+    },
+    {
+      "path": "../../../vendor/cordis"
+    },
+    {
+      "path": "../../core/agent"
+    },
+    {
+      "path": "../../llm/llm"
+    }
+  ]
+}

packages/util/README.md

@@ -5,5 +5,8 @@ Zero-dependency primitives shared across the other groups. A package lands here
 | Package | Role |
 |---|---|
 | `brand/` | The type-only `Branded<B>` nominal-typing primitive (no runtime code, no harness deps) |
+| `timeout/` | The timing/classification half of a timeout — `clampTimeout`/`deadline`/`timeoutOf`/`TimeoutReason` (pure functions, no harness deps); termination stays in each capability |
 
 `dsh-brand` is the canonical case: it owns ONLY the `Branded<B>` helper, so a capability package can brand the ids it owns (`dsh-tasks`'s `TaskId`, `dsh-session`'s `SessionId`, …) by depending on `dsh-brand` alone, without pulling in an unrelated package just to reach `Branded`.
+
+`dsh-timeout` follows the same shape for the timeout family: `dsh-bash` and `dsh-web-fetch-local` each fuse a caller's cancellation with a deadline and later classify "timed out" vs "cancelled" by depending on `dsh-timeout` alone. It deliberately owns only the timing/classification half — the *termination* (SIGKILL a process group, tear down a fetch socket) stays in each capability, because no shared layer can own every capability's kill (see [the timeout-library RFC](../../docs/rfc/implemented/architecture/2026-07-06-timeout-deadline-library.md)).

packages/util/timeout/README.md

@@ -0,0 +1,42 @@
+# dsh-timeout
+
+The **timing-and-classification** half of a timeout — a zero-dependency library of pure functions (no runtime harness deps) shared by every capability that clamps a caller's timeout hint, arms a deadline, and later has to tell "timed out" apart from "cancelled".
+
+It owns **no termination**. The signal it hands out only *notifies*; actually stopping the work stays in each capability, because that mechanism differs — bash SIGKILLs an OS process group, web tears down a `fetch` socket — and no shared layer can own all of them. This is the boundary the [RFC](../../../docs/rfc/implemented/architecture/2026-07-06-timeout-deadline-library.md) draws: share the timing/classification, keep the hard kill local.
+
+It is a **library, not a service or plugin**: no `ctx`, registers nothing, holds no state, emits no events. A "timeout service" would have to understand how to stop every capability's work — exactly the knowledge a microkernel keeps out of shared layers.
+
+## Surface
+
+```ts
+import { clampTimeout, deadline, timeoutOf, TimeoutReason } from '@deepseek-ai/dsh-timeout'
+```
+
+| Export | Role |
+|---|---|
+| `clampTimeout(requested, def, max, name?)` | Validate the caller's optional positive-finite hint, fill from `def`, cap at `max`. Throws (with `name`) on a non-positive/non-finite hint. |
+| `deadline(upstream, timeoutMs, code)` | Fuse `upstream` cancellation with a timeout into one `AbortSignal` (`AbortSignal.any`); the timeout carries a `TimeoutReason`. `[Symbol.dispose]` clears the timer. |
+| `timeoutOf(signal \| { reason }, code?)` | Recover the `TimeoutReason` from an aborted signal/error, else `undefined` — the timeout-vs-cancel classifier. Pass `code` to match only THIS deadline's timer (see nesting below). |
+| `TimeoutReason` | The internal reason (`code` + `timeoutMs`) stamped on a timeout abort. Not a public error — providers translate it into their own error/field. |
+
+## The `timeoutMs <= 0` sentinel
+
+`0` is the **internal** "no timeout" value for backend-owned background work (bash `start()`): `deadline()` arms no timer and forwards only `upstream`; with no upstream either, it returns a never-aborting signal plus a no-op disposer, so every caller keeps one call shape. External request hints validate as **positive finite** via `clampTimeout` before they reach `deadline`, so `0` is never a model-/plugin-facing "disable timeout" value.
+
+## Usage shape
+
+```ts ignore-check
+// Scope-lifetime consumer (foreground bash, one fetch): `using` disposes the timer.
+using d = deadline(upstream, timeoutMs, 'BASH_TIMEOUT')
+const outcome = await runWork({ signal: d.signal })            // work listens on d.signal and terminates itself
+const timedOut = timeoutOf(d.signal, 'BASH_TIMEOUT') !== undefined  // classify the first abort, scoped to OUR code
+const aborted = d.signal.aborted && !timedOut                  // mutually exclusive: timeout won, or cancel did
+```
+
+The signal only *notifies* — the caller MUST attach its own termination (`d.signal.addEventListener('abort', kill)`, or hand `d.signal` to `fetch`). Racing a promise against a timer would resolve the tool-call while the child process or socket leaks on; handing out a signal forces a real termination path to exist.
+
+Pass your own `code` to `timeoutOf` so classification composes under nesting: when the `upstream` you were handed is *itself* a deadline signal (a future `tools/execute` middleware arming a per-call deadline), `AbortSignal.any` preserves the outer `TimeoutReason` if the outer timer fires first. Scoping to your `code` makes a foreign timeout read as an ordinary upstream cancel — the correct classification from your capability's view — instead of your own timeout firing when your local timer never expired.
+
+## What does NOT get a timeout
+
+Local file `read`/`write`/`edit` take no `timeoutMs`: a syscall is best-effort-abortable at most, a timeout could not force `fsync`/`rename` to stop, and adding one would be an implicit default that violates explicit-over-implicit. See [`fs/`](../../fs/README.md).

packages/util/timeout/package.json

@@ -0,0 +1,30 @@
+{
+  "name": "@deepseek-ai/dsh-timeout",
+  "description": "Zero-dependency timeout/deadline primitive: clampTimeout, deadline, timeoutOf, TimeoutReason (timing + classification only, no termination)",
+  "version": "0.0.1",
+  "private": true,
+  "type": "module",
+  "main": "lib/index.js",
+  "types": "lib/types/index.d.ts",
+  "exports": {
+    ".": {
+      "types": "./lib/types/index.d.ts",
+      "default": "./lib/index.js"
+    },
+    "./src/*": "./src/*",
+    "./package.json": "./package.json"
+  },
+  "files": [
+    "lib/index.js",
+    "lib/types/**/*.d.ts",
+    "lib/types/**/*.d.ts.map",
+    "src"
+  ],
+  "license": "BSD-3-Clause",
+  "peerDependencies": {
+    "cordis": "^4.0.0-rc.6"
+  },
+  "devDependencies": {
+    "cordis": "^4.0.0-rc.6"
+  }
+}

packages/util/timeout/src/index.ts

@@ -0,0 +1,162 @@
+/**
+ * The timing-and-classification half of a timeout — a zero-dependency library
+ * of pure functions shared by every capability that clamps a caller's timeout
+ * hint, arms a deadline, and later has to tell "timed out" apart from
+ * "cancelled". It owns NO termination: the returned {@link deadline} signal only
+ * NOTIFIES; actually stopping the work (SIGKILL a process group, tear down a
+ * fetch socket, …) stays in each capability's implementation, because that
+ * mechanism differs per capability and no shared layer can own all of them.
+ *
+ * This is deliberately a library, not a cordis service or plugin: it takes no
+ * `ctx`, registers nothing, holds no cross-call state, and emits no events. A
+ * "timeout service" would have to understand how to stop every capability's
+ * work — exactly the knowledge a microkernel keeps out of shared layers.
+ *
+ * The four exports and their division of labor:
+ * - {@link clampTimeout} — validate a caller's optional positive hint, fill the
+ *   backend default, cap at the backend max (pure arithmetic + the shared
+ *   positive-finite request contract).
+ * - {@link deadline} — fuse upstream cancellation with a timeout into one
+ *   `AbortSignal`, the timeout carrying an identifiable {@link TimeoutReason};
+ *   `[Symbol.dispose]` clears the timer.
+ * - {@link timeoutOf} — classify an aborted signal (or error): a
+ *   {@link TimeoutReason} means the timeout fired, anything else (or nothing)
+ *   means it did not.
+ * - {@link TimeoutReason} — the internal classification reason; providers
+ *   translate it into their own public error/result shape before returning.
+ *
+ * @module @deepseek-ai/dsh-timeout
+ */
+
+/**
+ * The internal reason attached to a timeout abort so consumers can classify it
+ * after the fact. It carries the failing `code` (each capability's own string —
+ * `BASH_TIMEOUT`, `WEB_FETCH_TIMEOUT`, …) and the `timeoutMs` that elapsed.
+ *
+ * It is an INTERNAL classification reason, not a public error: providers
+ * translate it into their seam-specific error code or result field (via
+ * {@link timeoutOf}) before returning to callers. Native `AbortSignal.timeout()`
+ * yields a fixed `TimeoutError` indistinguishable across timeout kinds; this
+ * type is identifiable and carries the code/duration.
+ */
+export class TimeoutReason extends Error {
+  override name = 'TimeoutReason'
+
+  /**
+   * @param code Capability-owned timeout code (e.g. `BASH_TIMEOUT`).
+   * @param timeoutMs The deadline that elapsed, in milliseconds.
+   */
+  constructor(readonly code: string, readonly timeoutMs: number) {
+    super(`${code} after ${timeoutMs}ms`)
+  }
+}
+
+/**
+ * Validate a caller's optional timeout hint, fill it from the backend default,
+ * then cap at the backend max. The shared positive-finite request contract:
+ * a supplied `requested` must be a positive finite number or this throws —
+ * `0` is NOT a caller-facing "disable timeout" value (that sentinel is internal
+ * to {@link deadline}). A missing `requested` falls back to `def`.
+ *
+ * @param requested The caller's optional hint; validated when present.
+ * @param def The backend default applied when `requested` is absent.
+ * @param max The backend upper bound the result is capped to.
+ * @param name Field name used in the thrown message (so the caller sees which input was bad).
+ * @returns The effective timeout in milliseconds: `min(requested ?? def, max)`.
+ */
+export function clampTimeout(
+  requested: number | undefined,
+  def: number,
+  max: number,
+  name = 'timeoutMs',
+): number {
+  if (requested !== undefined && (!Number.isFinite(requested) || requested <= 0)) {
+    throw new Error(`${name} must be a positive finite number`)
+  }
+  return Math.min(requested ?? def, max)
+}
+
+/** A deadline signal plus the cleanup that clears its timer (dispose-once). */
+export interface Deadline {
+  /** Aborts on upstream cancellation OR on timeout (the timeout carries a {@link TimeoutReason}). */
+  readonly signal: AbortSignal
+  /** Clear the timer. Safe to call once; `using` calls it at scope exit. */
+  [Symbol.dispose](): void
+}
+
+/**
+ * Build a deadline signal that aborts on upstream cancellation OR on timeout,
+ * with the timeout carrying an identifiable {@link TimeoutReason} (unlike
+ * native `AbortSignal.timeout()`, whose fixed `TimeoutError` is opaque). It is
+ * `AbortSignal.any([upstream, <timeout>])` — the single primitive that fuses
+ * two abort sources — with the reason and a disposable timer added on top.
+ *
+ * `timeoutMs <= 0` is the INTERNAL "no timeout" sentinel for backend-owned
+ * background work: arm no timer and forward only the upstream signal; with no
+ * upstream either, return a never-aborting signal so callers keep one call
+ * shape. External request hints validate as positive finite via
+ * {@link clampTimeout} before reaching here, so `0` never arrives from a model
+ * or plugin.
+ *
+ * The returned object's `[Symbol.dispose]` clears the timer — use `using` for a
+ * scope-lifetime consumer, or call it manually for an event-lifetime one. The
+ * signal only NOTIFIES; the caller must attach its own termination (kill the
+ * process group, abort the fetch, …).
+ *
+ * @param upstream The caller's cancellation signal, if any, fused into the result.
+ * @param timeoutMs Deadline in milliseconds; `<= 0` means "no timeout" (arm no timer).
+ * @param code Capability-owned code stamped onto the timeout's {@link TimeoutReason}.
+ * @returns The fused {@link Deadline} (signal + timer cleanup).
+ */
+export function deadline(
+  upstream: AbortSignal | undefined,
+  timeoutMs: number,
+  code: string,
+): Deadline {
+  if (timeoutMs <= 0) {
+    // No timeout (background work): forward only the upstream signal, or a
+    // never-aborting one when there is no upstream. No timer, so dispose is a
+    // no-op — the empty method keeps the one call shape for every caller.
+    return { signal: upstream ?? new AbortController().signal, [Symbol.dispose]() {} }
+  }
+
+  const timer = new AbortController()
+  const id = setTimeout(() => { timer.abort(new TimeoutReason(code, timeoutMs)) }, timeoutMs)
+  return {
+    // AbortSignal.any adopts the reason of whichever source aborts FIRST, so a
+    // race resolves to a single cause: timeoutOf() reads TimeoutReason only
+    // when the timeout won, and upstream-wins leaves an ordinary abort reason.
+    signal: upstream !== undefined ? AbortSignal.any([upstream, timer.signal]) : timer.signal,
+    [Symbol.dispose]() { clearTimeout(id) },
+  }
+}
+
+/**
+ * Recover the {@link TimeoutReason} from an aborted signal (or any object with a
+ * `reason`), else `undefined`. This is the classification half: a provider
+ * calls it on the deadline signal after an abort to decide whether the cause
+ * was its timeout (translate to the capability's timeout error/field) or an
+ * ordinary upstream cancellation (`undefined` → the cancel path).
+ *
+ * Pass `code` to scope the match to THIS deadline's timer. It matters under
+ * nesting: when the `upstream` handed to {@link deadline} is itself a deadline
+ * signal (e.g. a future `tools/execute` middleware arming a per-call deadline),
+ * `AbortSignal.any` preserves the OUTER `TimeoutReason` if the outer timer fires
+ * first. Without `code`, the inner capability would misclassify that outer
+ * timeout as its own (`timedOut:true` / `WEB_FETCH_TIMEOUT`) though its local
+ * timer never expired; with `code`, a foreign timeout reads as `undefined` and
+ * falls through to the upstream-cancel path, which is the correct classification
+ * from the inner capability's view. Omit `code` only to ask "was this ANY
+ * timeout" (a generic middleware that owns no single code).
+ *
+ * @param x An {@link AbortSignal} or any `{ reason }` carrier (e.g. a caught abort error).
+ * @param code When provided, only a {@link TimeoutReason} with this exact `code` matches.
+ * @returns The matching {@link TimeoutReason}, else `undefined`.
+ */
+export function timeoutOf(x: AbortSignal | { reason?: unknown }, code?: string): TimeoutReason | undefined {
+  // AbortSignal.reason is typed `any`; pin it to `unknown` so no `any` leaks and
+  // the instanceof narrows cleanly for both a signal and a bare reason carrier.
+  const reason: unknown = x.reason
+  if (!(reason instanceof TimeoutReason)) return undefined
+  return code === undefined || reason.code === code ? reason : undefined
+}

packages/util/timeout/tests/timeout.spec.ts

@@ -0,0 +1,185 @@
+import { afterEach, describe, expect, it, vi } from 'vitest'
+import { clampTimeout, deadline, timeoutOf, TimeoutReason } from '@deepseek-ai/dsh-timeout'
+
+describe('TimeoutReason', () => {
+  it('is an Error carrying the code and elapsed ms', () => {
+    const reason = new TimeoutReason('BASH_TIMEOUT', 100)
+    expect(reason).toBeInstanceOf(Error)
+    expect(reason.name).toBe('TimeoutReason')
+    expect(reason.code).toBe('BASH_TIMEOUT')
+    expect(reason.timeoutMs).toBe(100)
+    expect(reason.message).toBe('BASH_TIMEOUT after 100ms')
+  })
+})
+
+describe('clampTimeout', () => {
+  it('fills the default when the hint is absent', () => {
+    expect(clampTimeout(undefined, 120_000, 600_000)).toBe(120_000)
+  })
+
+  it('caps the hint at max', () => {
+    expect(clampTimeout(999_999, 120_000, 600_000)).toBe(600_000)
+  })
+
+  it('keeps a valid hint under the cap', () => {
+    expect(clampTimeout(5_000, 120_000, 600_000)).toBe(5_000)
+  })
+
+  it('caps the default itself when the default exceeds max', () => {
+    // min(def, max) applies even with no hint — a misconfigured backend never
+    // exceeds its own cap.
+    expect(clampTimeout(undefined, 900_000, 600_000)).toBe(600_000)
+  })
+
+  it('rejects a non-finite hint with the caller-provided name', () => {
+    expect(() => clampTimeout(Number.NaN, 100, 200, 'bash-local: request.timeoutMs'))
+      .toThrow(/bash-local: request\.timeoutMs must be a positive finite number/)
+    expect(() => clampTimeout(Number.POSITIVE_INFINITY, 100, 200))
+      .toThrow(/timeoutMs must be a positive finite number/)
+  })
+
+  it('rejects a non-positive hint', () => {
+    expect(() => clampTimeout(0, 100, 200)).toThrow(/must be a positive finite number/)
+    expect(() => clampTimeout(-1, 100, 200)).toThrow(/must be a positive finite number/)
+  })
+})
+
+describe('deadline — timeout arm', () => {
+  afterEach(() => { vi.useRealTimers() })
+
+  it('aborts on timeout with a TimeoutReason after the elapsed ms', () => {
+    vi.useFakeTimers()
+    using d = deadline(undefined, 100, 'BASH_TIMEOUT')
+    expect(d.signal.aborted).toBe(false)
+    vi.advanceTimersByTime(100)
+    expect(d.signal.aborted).toBe(true)
+    const reason = timeoutOf(d.signal)
+    expect(reason).toBeInstanceOf(TimeoutReason)
+    expect(reason?.code).toBe('BASH_TIMEOUT')
+    expect(reason?.timeoutMs).toBe(100)
+  })
+
+  it('[Symbol.dispose] clears the timer so no abort fires afterward', () => {
+    vi.useFakeTimers()
+    const d = deadline(undefined, 100, 'BASH_TIMEOUT')
+    d[Symbol.dispose]()
+    vi.advanceTimersByTime(1_000)
+    expect(d.signal.aborted).toBe(false)
+    expect(timeoutOf(d.signal)).toBeUndefined()
+  })
+})
+
+describe('deadline — fuse with upstream', () => {
+  it('aborts on upstream cancellation, classified as NOT a timeout', () => {
+    const upstream = new AbortController()
+    using d = deadline(upstream.signal, 60_000, 'BASH_TIMEOUT')
+    upstream.abort('user cancelled')
+    expect(d.signal.aborted).toBe(true)
+    expect(timeoutOf(d.signal)).toBeUndefined()
+  })
+
+  it('cancel wins when it fires before the timeout', () => {
+    vi.useFakeTimers()
+    try {
+      const upstream = new AbortController()
+      using d = deadline(upstream.signal, 100, 'BASH_TIMEOUT')
+      upstream.abort('user cancelled') // fires first, before the 100ms timer
+      vi.advanceTimersByTime(200)
+      expect(d.signal.aborted).toBe(true)
+      // AbortSignal.any adopts the FIRST source's reason: cancel won, so no
+      // TimeoutReason even though the timer later elapsed.
+      expect(timeoutOf(d.signal)).toBeUndefined()
+    } finally {
+      vi.useRealTimers()
+    }
+  })
+
+  it('timeout wins when it fires before upstream cancellation', () => {
+    vi.useFakeTimers()
+    try {
+      const upstream = new AbortController()
+      using d = deadline(upstream.signal, 100, 'WEB_FETCH_TIMEOUT')
+      vi.advanceTimersByTime(150) // past the 100ms deadline: the timer fires first
+      expect(d.signal.aborted).toBe(true)
+      expect(timeoutOf(d.signal)?.code).toBe('WEB_FETCH_TIMEOUT')
+      // A later upstream abort is a no-op on the already-aborted fused signal:
+      // AbortSignal.any keeps the FIRST cause, so the timeout classification stands.
+      upstream.abort('too late')
+      expect(timeoutOf(d.signal)?.code).toBe('WEB_FETCH_TIMEOUT')
+    } finally {
+      vi.useRealTimers()
+    }
+  })
+
+  it('forwards a pre-aborted upstream signal immediately', () => {
+    const upstream = new AbortController()
+    upstream.abort('already gone')
+    using d = deadline(upstream.signal, 60_000, 'BASH_TIMEOUT')
+    expect(d.signal.aborted).toBe(true)
+    expect(timeoutOf(d.signal)).toBeUndefined()
+  })
+})
+
+describe('deadline — timeoutMs <= 0 (no-timeout sentinel)', () => {
+  afterEach(() => { vi.useRealTimers() })
+
+  it('arms no timer and forwards only the upstream signal', () => {
+    vi.useFakeTimers()
+    const upstream = new AbortController()
+    using d = deadline(upstream.signal, 0, 'BASH_TIMEOUT')
+    vi.advanceTimersByTime(1_000_000)
+    expect(d.signal.aborted).toBe(false) // no timer ever armed
+    upstream.abort('kill')
+    expect(d.signal.aborted).toBe(true)
+    expect(timeoutOf(d.signal)).toBeUndefined() // never a timeout
+  })
+
+  it('returns a never-aborting signal with a no-op disposer when there is no upstream', () => {
+    vi.useFakeTimers()
+    const d = deadline(undefined, 0, 'BASH_TIMEOUT')
+    expect(() => { d[Symbol.dispose]() }).not.toThrow()
+    vi.advanceTimersByTime(1_000_000)
+    expect(d.signal.aborted).toBe(false)
+    expect(timeoutOf(d.signal)).toBeUndefined()
+  })
+
+  it('treats a negative timeout the same as zero', () => {
+    const d = deadline(undefined, -5, 'BASH_TIMEOUT')
+    expect(d.signal.aborted).toBe(false)
+    d[Symbol.dispose]()
+  })
+})
+
+describe('timeoutOf', () => {
+  it('classifies a bare reason carrier that holds a TimeoutReason', () => {
+    const reason = new TimeoutReason('WEB_FETCH_TIMEOUT', 50)
+    expect(timeoutOf({ reason })).toBe(reason)
+  })
+
+  it('returns undefined for a non-timeout reason', () => {
+    expect(timeoutOf({ reason: new Error('other') })).toBeUndefined()
+    expect(timeoutOf({ reason: 'user cancelled' })).toBeUndefined()
+    expect(timeoutOf({})).toBeUndefined()
+  })
+
+  it('matches only the requested code when one is given', () => {
+    const reason = new TimeoutReason('BASH_TIMEOUT', 100)
+    expect(timeoutOf({ reason }, 'BASH_TIMEOUT')).toBe(reason)
+    expect(timeoutOf({ reason }, 'WEB_FETCH_TIMEOUT')).toBeUndefined()
+  })
+})
+
+describe('deadline — nested deadlines', () => {
+  it("does not misclassify an outer deadline's timeout as the inner code", () => {
+    // The upstream handed to the inner deadline is ITSELF a deadline that has
+    // already timed out (outer). AbortSignal.any preserves the outer reason;
+    // scoping timeoutOf to the inner code keeps the inner capability from
+    // reporting the outer timeout as its own — it reads as an upstream cancel.
+    const outer = new AbortController()
+    outer.abort(new TimeoutReason('OUTER_TIMEOUT', 30))
+    using inner = deadline(outer.signal, 60_000, 'BASH_TIMEOUT')
+    expect(inner.signal.aborted).toBe(true)
+    expect(timeoutOf(inner.signal, 'BASH_TIMEOUT')).toBeUndefined() // not ours → upstream-cancel path
+    expect(timeoutOf(inner.signal)?.code).toBe('OUTER_TIMEOUT') // but IS a timeout, unscoped
+  })
+})

packages/util/timeout/tsconfig.json

@@ -0,0 +1,11 @@
+{
+  "extends": "../../../tsconfig.base.json",
+  "compilerOptions": {
+    "rootDir": "src",
+    "outDir": "lib/types"
+  },
+  "include": [
+    "src"
+  ],
+  "references": []
+}

packages/web/tool-web/README.md

@@ -1,6 +1,6 @@
 # @deepseek-ai/dsh-tool-web
 
-The model-facing web tool suite — `web_search` and `web_fetch` — over the [web capability seam](../web/README.md) (`ctx.web`). It owns model-facing concerns only: tool names, JSON schemas, snake_case argument names, prompt sections, the result-count bound, result formatting, HTML→markdown presentation, and `presentCall`. All web access goes through `ctx.web`; this package never imports a concrete provider.
+The model-facing web tool suite — `web_search` and `web_fetch` — over the [web capability seam](../web/README.md) (`ctx.web`). It owns model-facing concerns only: tool names, JSON schemas, snake_case argument names, prompt sections, the result-count bound, result formatting, HTML→markdown presentation, and `presentCall`. All web access goes through `ctx.web`; this package never imports a concrete provider. Neither tool exposes a model-facing timeout — each tool's cooperative tool-call budget is declared here via config (`fetchTimeoutMs`/`searchTimeoutMs`, attached as `ToolDefinition.timeoutMs`) and enforced by [`@deepseek-ai/dsh-timeout-policy`](../../timeout/timeout-policy/README.md) (a `tools/execute` wrapper); each tool just forwards `exec.signal` to the seam.
 
 Each tool is registered independently; a product that wants only one disables the other via config (`{ search: false }` / `{ fetch: false }`).
 
@@ -9,7 +9,7 @@ Each tool is registered independently; a product that wants only one disables th
 | Tool | Args | Behavior |
 |---|---|---|
 | `web_search` | `query` (string) | Discovery. Returns an optional answer plus source URLs. `max_results` is **not** model-facing — the tool sets the bound (the `searchMaxResults` config, default 8) and passes it to the seam. |
-| `web_fetch` | `url` (string), `timeout_ms` (number, optional) | Retrieves a specific URL. HTML bodies are rendered to markdown-ish text; text bodies pass through. A non-2xx status is reported, not an error. |
+| `web_fetch` | `url` (string) | Retrieves a specific URL. HTML bodies are rendered to markdown-ish text; text bodies pass through. A non-2xx status is reported, not an error. The tool-call timeout is deployment policy (`dsh-timeout-policy`), not a model argument. |
 
 ## Config
 
@@ -18,6 +18,10 @@ Each tool is registered independently; a product that wants only one disables th
 | `search` | `true` | Register `web_search`. |
 | `fetch` | `true` | Register `web_fetch`. |
 | `searchMaxResults` | `8` | Upper bound on sources returned by one `web_search` call (the seam truncates a longer provider list and flags it). |
+| `fetchTimeoutMs` | `30000` | Cooperative tool-call timeout budget (ms) for `web_fetch`. |
+| `searchTimeoutMs` | `30000` | Cooperative tool-call timeout budget (ms) for `web_search`. |
+
+`fetchTimeoutMs`/`searchTimeoutMs` declare each tool's cooperative timeout budget (attached as `ToolDefinition.timeoutMs`), enforced by [`@deepseek-ai/dsh-timeout-policy`](../../timeout/timeout-policy/README.md); the model-facing schema exposes no timeout argument.
 
 ```yaml
 - id: tool-web

packages/web/tool-web/package.json

@@ -37,6 +37,7 @@
     "@deepseek-ai/dsh-session": "workspace:^",
     "@deepseek-ai/dsh-system-prompt": "workspace:^",
     "@deepseek-ai/dsh-tools": "workspace:^",
+    "@deepseek-ai/dsh-timeout-policy": "workspace:^",
     "@deepseek-ai/dsh-web": "workspace:^",
     "@deepseek-ai/dsh-web-fetch-local": "workspace:^",
     "@deepseek-ai/dsh-web-search-exa": "workspace:^",

packages/web/tool-web/src/fetch.ts

@@ -3,6 +3,13 @@
  * Execution goes through `ctx.web` — this module owns the model-facing schema,
  * argument validation, and PRESENTATION (HTML→markdown, truncation formatting),
  * while the fetch provider owns safe retrieval (transport, redirects, caps).
+ *
+ * The model-facing schema exposes NO timeout knob: the tool-call budget is
+ * deployment policy DECLARED via this package's `fetchTimeoutMs` config (attached
+ * as `ToolDefinition.timeoutMs`) and ENFORCED by `@deepseek-ai/dsh-timeout-policy`
+ * (a `tools/execute` wrapper), matching the reference-agent `WebFetch` shape. This
+ * tool just forwards the (possibly deadline-derived) `exec.signal` to `ctx.web`;
+ * the provider keeps its own timeout only as a resource backstop for direct callers.
  */
 
 import type { Context } from 'cordis'
@@ -15,18 +22,17 @@ import type {} from '@deepseek-ai/dsh-system-prompt'
 import { htmlToMarkdown } from './html.ts'
 
 /**
- * Validate value constraints the schema DSL can't express: a non-blank `url`,
- * and a positive `timeout_ms` when present. Throws a plain `Error` otherwise.
+ * Validate value constraints the schema DSL can't express: a non-blank `url`.
+ * Throws a plain `Error` otherwise. No timeout parameter — the tool-call budget
+ * is deployment policy declared via `fetchTimeoutMs` config and enforced by
+ * `@deepseek-ai/dsh-timeout-policy`, not a model argument.
  *
  * @param args - the schema-validated `web_fetch` arguments.
- * @returns the arguments renamed to the seam's camelCase request fields.
+ * @returns the arguments as the seam's request fields.
  */
-export function parseFetchArgs(args: { url: string; timeout_ms?: number }): { url: string; timeoutMs?: number } {
+export function parseFetchArgs(args: { url: string }): { url: string } {
   if (args.url.trim().length === 0) throw new Error('url must be a non-empty string')
-  if (args.timeout_ms !== undefined && (!Number.isFinite(args.timeout_ms) || args.timeout_ms <= 0)) {
-    throw new Error('timeout_ms must be a positive number')
-  }
-  return { url: args.url, ...args.timeout_ms !== undefined ? { timeoutMs: args.timeout_ms } : {} }
+  return { url: args.url }
 }
 
 /**
@@ -67,7 +73,7 @@ export function formatFetchOutput(result: WebFetchResult): string {
  * @param args - the raw tool arguments; only `url` feeds the view.
  * @returns the generic card view (`kind: 'fetch'`) shown while the call runs.
  */
-export function presentFetchCall(args: { url: string; timeout_ms?: number }): GenericCallView {
+export function presentFetchCall(args: { url: string }): GenericCallView {
   return { card: 'generic', title: args.url, kind: 'fetch', rawInput: args.url }
 }
 
@@ -76,8 +82,10 @@ export function presentFetchCall(args: { url: string; timeout_ms?: number }): Ge
  *
  * @param ctx - context whose `tools` and `systemPrompt` registries receive the
  *   registrations; both are effect-scoped and unregister on plugin dispose.
+ * @param timeoutMs - the cooperative tool-call budget (ms) attached as the tool's
+ *   `ToolDefinition.timeoutMs` for `@deepseek-ai/dsh-timeout-policy` to enforce.
  */
-export function applyWebFetchTool(ctx: Context): void {
+export function applyWebFetchTool(ctx: Context, timeoutMs: number): void {
   ctx.systemPrompt.section({
     name: 'tool:web_fetch',
     order: 111,
@@ -89,12 +97,12 @@ export function applyWebFetchTool(ctx: Context): void {
     description: 'Fetch the content of a specific HTTP(S) URL and return it decoded to text.',
     parameters: {
       url: { type: 'string', required: true, description: 'The HTTP(S) URL to fetch.' },
-      timeout_ms: { type: 'number', description: 'Optional fetch timeout in milliseconds (capped by the provider).' },
     },
+    timeoutMs,
     async execute(args, exec): Promise<ContentBlock[]> {
       const input = parseFetchArgs(args)
       const result = await ctx.web.fetch(
-        { url: input.url, ...input.timeoutMs !== undefined ? { timeoutMs: input.timeoutMs } : {} },
+        { url: input.url },
         exec.signal ? { signal: exec.signal } : undefined,
       )
       return [{ type: 'text', text: formatFetchOutput(result) }]

packages/web/tool-web/src/index.ts

@@ -33,7 +33,10 @@ export const name = 'tool-web'
 /** Services required by the web tool suite. */
 export const inject = ['tools', 'web', 'systemPrompt']
 
-/** Plugin config: which web tools to register, and the `web_search` source cap. */
+/** Default cooperative tool-call timeout budget (ms) for the web tools. */
+export const DEFAULT_WEB_TOOL_TIMEOUT_MS = 30_000
+
+/** Plugin config: which web tools to register, the source cap, and per-tool budgets. */
 export interface Config {
   /** Register `web_search`. Defaults to true. */
   search?: boolean
@@ -41,12 +44,18 @@ export interface Config {
   fetch?: boolean
   /** Upper bound on sources returned by one `web_search` call. */
   searchMaxResults?: number
+  /** Cooperative timeout budget (ms) for `web_fetch`. Defaults to 30000. */
+  fetchTimeoutMs?: number
+  /** Cooperative timeout budget (ms) for `web_search`. Defaults to 30000. */
+  searchTimeoutMs?: number
 }
 
 export const Config: z<Config> = z.object({
   search: z.boolean().default(true),
   fetch: z.boolean().default(true),
   searchMaxResults: z.number().default(WEB_SEARCH_MAX_RESULTS),
+  fetchTimeoutMs: z.number().default(DEFAULT_WEB_TOOL_TIMEOUT_MS),
+  searchTimeoutMs: z.number().default(DEFAULT_WEB_TOOL_TIMEOUT_MS),
 })
 
 /** The shape after schemastery applies its defaults to every field. */
@@ -61,7 +70,10 @@ function assertPositiveInteger(name: string, value: number): void {
 
 /**
  * Register the enabled web tools. `search`/`fetch` default to true; a product
- * that wants only one disables the other in config. The tools' disposers are
+ * that wants only one disables the other in config. Each tool's cooperative
+ * timeout budget (`fetchTimeoutMs`/`searchTimeoutMs`, default 30000) is resolved
+ * here and attached to the tool as `ToolDefinition.timeoutMs` for
+ * `@deepseek-ai/dsh-timeout-policy` to enforce. The tools' disposers are
  * fiber-scoped (the effect-based registries clean up on dispose), so no manual
  * teardown is needed.
  */
@@ -69,6 +81,8 @@ export function apply(ctx: Context, config: Config): void {
   // schemastery (Config) has already filled every defaulted field.
   const resolved = config as ResolvedConfig
   assertPositiveInteger('searchMaxResults', resolved.searchMaxResults)
-  if (resolved.search) applyWebSearchTool(ctx, resolved.searchMaxResults)
-  if (resolved.fetch) applyWebFetchTool(ctx)
+  assertPositiveInteger('fetchTimeoutMs', resolved.fetchTimeoutMs)
+  assertPositiveInteger('searchTimeoutMs', resolved.searchTimeoutMs)
+  if (resolved.search) applyWebSearchTool(ctx, resolved.searchMaxResults, resolved.searchTimeoutMs)
+  if (resolved.fetch) applyWebFetchTool(ctx, resolved.fetchTimeoutMs)
 }

packages/web/tool-web/src/search.ts

@@ -92,8 +92,10 @@ export function presentSearchCall(args: { query: string }): GenericCallView {
  *   registrations; both are effect-scoped and unregister on plugin dispose.
  * @param maxResults - the deployment's source cap, sent as every seam
  *   request's `maxResults`.
+ * @param timeoutMs - the cooperative tool-call budget (ms) attached as the tool's
+ *   `ToolDefinition.timeoutMs` for `@deepseek-ai/dsh-timeout-policy` to enforce.
  */
-export function applyWebSearchTool(ctx: Context, maxResults: number): void {
+export function applyWebSearchTool(ctx: Context, maxResults: number, timeoutMs: number): void {
   ctx.systemPrompt.section({
     name: 'tool:web_search',
     order: 110,
@@ -106,6 +108,7 @@ export function applyWebSearchTool(ctx: Context, maxResults: number): void {
     parameters: {
       query: { type: 'string', required: true, description: 'The search query.' },
     },
+    timeoutMs,
     async execute(args, exec): Promise<ContentBlock[]> {
       const input = parseSearchArgs(args)
       const result = await ctx.web.search(

packages/web/tool-web/tests/integration.spec.ts

@@ -1,10 +1,11 @@
 /**
  * Integration: the real fetch backend (`dsh-web-fetch-local`) + a real search
  * provider (`dsh-web-search-exa`) + the real seam (`dsh-web`) + the model tool
- * (`dsh-tool-web`), exercised through `ctx.tools.execute()` — nothing bypasses
- * the tool registry. Fetch hits a real loopback HTTP server (verifying the
- * WORLD); search runs the real Exa provider over a stubbed global `fetch` (the
- * network is the one boundary we mock).
+ * (`dsh-tool-web`) + the tool-call timeout policy (`dsh-timeout-policy`),
+ * exercised through `ctx.tools.execute()` — nothing bypasses the tool registry.
+ * Fetch hits a real loopback HTTP server (verifying the WORLD); search runs the
+ * real Exa provider over a stubbed global `fetch` (the network is the one
+ * boundary we mock).
  */
 
 import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
@@ -18,6 +19,7 @@ import WebService from '@deepseek-ai/dsh-web'
 import * as WebFetchLocal from '@deepseek-ai/dsh-web-fetch-local'
 import * as WebSearchExa from '@deepseek-ai/dsh-web-search-exa'
 import * as ToolWeb from '@deepseek-ai/dsh-tool-web'
+import * as TimeoutPolicy from '@deepseek-ai/dsh-timeout-policy'
 
 type Handler = (req: IncomingMessage, res: ServerResponse) => void
 
@@ -39,6 +41,11 @@ beforeEach(async () => {
   await ctx.plugin(WebService, { searchProvider: WebSearchExa.EXA_PROVIDER_ID, fetchProvider: WebFetchLocal.LOCAL_FETCH_PROVIDER_ID })
   await ctx.plugin(WebFetchLocal, {})
   await ctx.plugin(WebSearchExa, { apiKey: 'exa-key', baseURL: 'https://api.exa.test' })
+  // The shipped deployment shape: the tool-call budget is declared by tool-web
+  // config (default 30s, attached as ToolDefinition.timeoutMs) and enforced by
+  // the zero-config timeout-policy plugin, set above the provider backstop so the
+  // policy normally wins.
+  await ctx.plugin(TimeoutPolicy)
   fiber = await ctx.plugin(ToolWeb)
 })
 
@@ -96,3 +103,70 @@ describe('web_search integration over the real Exa provider', () => {
     expect(out.content.map(b => b.text).join('')).toContain('[Result](https://result.test)')
   })
 })
+
+describe('tool-call timeout policy over the migrated web tools', () => {
+  it('neither model schema exposes a timeout parameter after the migration', () => {
+    const byName = new Map(ctx.tools.schemas().map(s => [s.name, s]))
+    const fetchParams = byName.get('web_fetch')!.parameters as { properties: Record<string, unknown> }
+    const searchParams = byName.get('web_search')!.parameters as { properties: Record<string, unknown> }
+    expect(Object.keys(fetchParams.properties)).toEqual(['url'])
+    expect('timeout_ms' in fetchParams.properties).toBe(false)
+    expect(Object.keys(searchParams.properties)).toEqual(['query'])
+  })
+})
+
+describe('tool-call timeout returns TOOL_TIMEOUT (deadline wins over a slow fetch)', () => {
+  let slowServer: Server
+  let slowBase: string
+  let openSockets: ServerResponse[]
+  let tctx: Context
+  let tfiber: Awaited<ReturnType<Context['plugin']>>
+
+  beforeEach(async () => {
+    // A server that never responds: it holds the connection open until the
+    // client aborts. The cooperative deadline (via exec.signal → the fetch
+    // provider → undici) is what ends the call.
+    openSockets = []
+    slowServer = createServer((_req, res) => { openSockets.push(res) })
+    await new Promise<void>(resolve => slowServer.listen(0, '127.0.0.1', resolve))
+    slowBase = `http://127.0.0.1:${(slowServer.address() as AddressInfo).port}`
+
+    tctx = new Context()
+    await tctx.plugin(SystemPrompt)
+    await tctx.plugin(ToolRegistry)
+    await tctx.plugin(WebService, { fetchProvider: WebFetchLocal.LOCAL_FETCH_PROVIDER_ID })
+    // Provider backstop well ABOVE the tool-call budget, so the policy wins.
+    await tctx.plugin(WebFetchLocal, { timeoutMs: 30_000, maxTimeoutMs: 60_000 })
+    await tctx.plugin(TimeoutPolicy)
+    // The tool-call budget is declared by tool-web config, enforced by the policy.
+    tfiber = await tctx.plugin(ToolWeb, { fetchTimeoutMs: 50 })
+  })
+
+  afterEach(async () => {
+    for (const res of openSockets) res.destroy()
+    await tfiber.dispose()
+    await new Promise<void>(resolve => slowServer.close(() => { resolve() }))
+  })
+
+  it('returns a structured TOOL_TIMEOUT (not the provider WEB_FETCH_TIMEOUT) when the tool-call budget wins', async () => {
+    const out = await tctx.tools.execute({ callId: CallId('slow-1'), name: 'web_fetch', arguments: { url: slowBase } })
+    expect(out.isError).toBe(true)
+    // The outer tool-call deadline won: TOOL_TIMEOUT, owned by dsh-timeout-policy,
+    // NOT the provider's own WEB_FETCH_TIMEOUT (its 30s backstop never fired).
+    expect(out.error?.code).toBe('TOOL_TIMEOUT')
+    const text = out.content.map(b => (b.type === 'text' ? b.text : '')).join('')
+    expect(text).toContain('timed out after 50ms')
+  })
+
+  it('the provider backstop still protects a DIRECT ctx.web.fetch() call (no tool-call policy in that path)', async () => {
+    // A direct seam caller does not go through tools/execute, so the tool-call
+    // policy never applies; the provider's OWN timeout is the only budget. A
+    // short per-request hint proves the provider backstop is intact and classifies
+    // as WEB_FETCH_TIMEOUT (the provider-owned code), never TOOL_TIMEOUT.
+    const err = await tctx.web.fetch({ url: slowBase, timeoutMs: 50 }).then(
+      () => undefined,
+      (e: unknown) => e as { code?: string },
+    )
+    expect(err?.code).toBe('WEB_FETCH_TIMEOUT')
+  })
+})

packages/web/tool-web/tests/tool-web.spec.ts

@@ -110,10 +110,9 @@ describe('fetch formatting', () => {
     expect(renderBody({ kind: 'html', content: '<p>y</p>' })).toBe('y')
   })
 
-  it('validates url and timeout', () => {
+  it('validates url (non-empty), no timeout parameter', () => {
     expect(() => parseFetchArgs({ url: ' ' })).toThrow('non-empty')
-    expect(() => parseFetchArgs({ url: 'https://a.test', timeout_ms: -1 })).toThrow('positive')
-    expect(parseFetchArgs({ url: 'https://a.test', timeout_ms: 5 })).toEqual({ url: 'https://a.test', timeoutMs: 5 })
+    expect(parseFetchArgs({ url: 'https://a.test' })).toEqual({ url: 'https://a.test' })
   })
 
   it('presents a fetch call as a fetch-kind card titled by the url', () => {
@@ -249,7 +248,7 @@ describe('tool-web execution through the real registry', () => {
     expect('default' in ToolWeb).toBe(false)
   })
 
-  it('executes web_fetch, forwarding timeout_ms and the abort signal to the seam', async () => {
+  it('executes web_fetch, forwarding the url (no timeout param) and the abort signal to the seam', async () => {
     const seen: { request?: { url: string; timeoutMs?: number }; signal?: AbortSignal | undefined } = {}
     const fetchProvider = {
       id: 'stub-fetch',
@@ -262,13 +261,35 @@ describe('tool-web execution through the real registry', () => {
     }
     const { ctx, fiber } = await mountTools({ webConfig: { fetchProvider: 'stub-fetch' }, fetchProvider })
     const controller = new AbortController()
-    const out = await ctx.tools.execute({ callId: CallId('fetch-1'), name: 'web_fetch', arguments: { url: 'https://a.test', timeout_ms: 1234 }, signal: controller.signal })
+    const out = await ctx.tools.execute({ callId: CallId('fetch-1'), name: 'web_fetch', arguments: { url: 'https://a.test' }, signal: controller.signal })
     expect(out.isError).toBe(false)
-    expect(seen.request).toEqual({ url: 'https://a.test', timeoutMs: 1234 })
+    // The model schema exposes no timeout: the tool forwards only the url; the
+    // tool-call budget is owned by dsh-timeout-policy over exec.signal.
+    expect(seen.request).toEqual({ url: 'https://a.test' })
     expect(seen.signal).toBe(controller.signal)
     await fiber.dispose()
   })
 
+  it('executes web_fetch with no caller signal (forwards undefined to the seam)', async () => {
+    const seen: { signal?: AbortSignal | undefined; passedExec?: boolean } = {}
+    const fetchProvider = {
+      id: 'stub-fetch',
+      status: () => available,
+      fetch: (request: { url: string }, exec?: { signal?: AbortSignal }) => {
+        seen.passedExec = exec !== undefined
+        seen.signal = exec?.signal
+        return Promise.resolve({ providerId: 'stub-fetch', url: request.url, statusCode: 200, body: { kind: 'text' as const, content: 'ok' }, truncated: false })
+      },
+    }
+    const { ctx, fiber } = await mountTools({ webConfig: { fetchProvider: 'stub-fetch' }, fetchProvider })
+    // No signal on the execution: the tool passes `undefined` (not `{ signal: undefined }`).
+    const out = await ctx.tools.execute({ callId: CallId('fetch-2'), name: 'web_fetch', arguments: { url: 'https://a.test' } })
+    expect(out.isError).toBe(false)
+    expect(seen.passedExec).toBe(false)
+    expect(seen.signal).toBeUndefined()
+    await fiber.dispose()
+  })
+
   it('executes web_search, forwarding the abort signal to the seam', async () => {
     const seen: { signal?: AbortSignal | undefined } = {}
     const provider: WebSearchProvider = {
@@ -328,3 +349,31 @@ describe('searchMaxResults is plugin config', () => {
       .rejects.toThrow(/tool-web: searchMaxResults must be a positive integer/)
   })
 })
+
+describe('tool-call timeout budget is plugin config', () => {
+  it('attaches the default 30s budget to web_fetch and web_search', async () => {
+    const { fiber, ctx } = await mountTools()
+    expect(ctx.tools.get('web_fetch')?.timeoutMs).toBe(30_000)
+    expect(ctx.tools.get('web_search')?.timeoutMs).toBe(30_000)
+    await fiber.dispose()
+  })
+
+  it('honors per-tool timeout overrides from config', async () => {
+    const { fiber, ctx } = await mountTools({ config: { fetchTimeoutMs: 60_000, searchTimeoutMs: 10_000 } })
+    expect(ctx.tools.get('web_fetch')?.timeoutMs).toBe(60_000)
+    expect(ctx.tools.get('web_search')?.timeoutMs).toBe(10_000)
+    await fiber.dispose()
+  })
+
+  it.each([
+    ['fetchTimeoutMs', { fetchTimeoutMs: 0 }],
+    ['searchTimeoutMs', { searchTimeoutMs: -5 }],
+  ])('rejects a non-positive-integer %s at load', async (key, config) => {
+    const ctx = new Context()
+    await ctx.plugin(SystemPrompt)
+    await ctx.plugin(ToolRegistry)
+    await ctx.plugin(WebService, {})
+    await expect(ctx.plugin(ToolWeb, config))
+      .rejects.toThrow(new RegExp(`tool-web: ${key} must be a positive integer`))
+  })
+})

packages/web/tool-web/tsconfig.json

@@ -12,6 +12,7 @@
     { "path": "../../llm/llm" },
     { "path": "../../core/tools" },
     { "path": "../../core/system-prompt" },
+    { "path": "../../timeout/timeout-policy" },
     { "path": "../web" }
   ]
 }

packages/web/web-fetch-local/README.md

@@ -6,7 +6,9 @@ This is an **implementation** package: it registers a provider into `ctx.web`, i
 
 ## Responsibility split
 
-The provider owns **safe resource retrieval**: URL validation, HTTP transport, redirect policy, timeout, abort propagation, byte caps, charset decoding, content-type classification, and binary rejection. `@deepseek-ai/dsh-tool-web` owns **presentation** (HTML→markdown, truncation formatting). A non-2xx HTTP response is a *result* (status code + decoded body), not an error; `WebError` is reserved for failures to safely retrieve or represent the resource.
+The provider owns **safe resource retrieval**: URL validation, HTTP transport, redirect policy, a resource-backstop timeout, abort propagation, byte caps, charset decoding, content-type classification, and binary rejection. `@deepseek-ai/dsh-tool-web` owns **presentation** (HTML→markdown, truncation formatting). A non-2xx HTTP response is a *result* (status code + decoded body), not an error; `WebError` is reserved for failures to safely retrieve or represent the resource.
+
+The provider's `timeoutMs`/`maxTimeoutMs` is a **resource backstop** for direct `ctx.web.fetch()` callers and misconfigured deployments — it is NOT the model-facing tool-call budget. The tool-call budget for `web_fetch` is deployment policy owned by [`@deepseek-ai/dsh-timeout-policy`](../../timeout/timeout-policy/README.md), which arms a per-call deadline on `exec.signal`. A shipped web-tool deployment sets the provider backstop **above** the `tool-timeout` budget, so the tool-call policy normally wins for model calls (returning `TOOL_TIMEOUT`); when the outer deadline signal reaches this provider first, it classifies as `WEB_ABORTED` and the outer wrapper replaces the result with `TOOL_TIMEOUT`. The provider's own `WEB_FETCH_TIMEOUT` only fires for a direct seam caller whose own budget elapsed.
 
 ## Transport hygiene
 
@@ -24,8 +26,8 @@ The provider owns **safe resource retrieval**: URL validation, HTTP transport, r
 | `maxUrlLength` | `2048` | Maximum accepted request URL length. |
 | `maxResponseBytes` | `5_000_000` | Maximum response body size in bytes. |
 | `maxBodyChars` | `100_000` | Maximum decoded body length in characters. |
-| `timeoutMs` | `30_000` | Default fetch timeout. |
-| `maxTimeoutMs` | `120_000` | Upper bound for a per-request timeout override. |
+| `timeoutMs` | `30_000` | Default fetch timeout — a resource backstop for direct `ctx.web.fetch()` callers, not the model-facing tool-call budget (that is `dsh-timeout-policy`). |
+| `maxTimeoutMs` | `120_000` | Upper bound for a per-request timeout override (direct callers). |
 | `maxRedirects` | `5` | Maximum same-origin redirect hops (`0` follows none). |
 | `userAgent` | `deepseek-harness/…` | `User-Agent` header. |
 

packages/web/web-fetch-local/package.json

@@ -22,6 +22,7 @@
   ],
   "license": "BSD-3-Clause",
   "peerDependencies": {
+    "@deepseek-ai/dsh-timeout": "^0.0.1",
     "@deepseek-ai/dsh-web": "^0.0.1",
     "cordis": "^4.0.0-rc.6"
   },
@@ -29,6 +30,7 @@
     "schemastery": "^3.18.0"
   },
   "devDependencies": {
+    "@deepseek-ai/dsh-timeout": "workspace:^",
     "@deepseek-ai/dsh-web": "workspace:^",
     "cordis": "^4.0.0-rc.6"
   }

packages/web/web-fetch-local/src/provider.ts

@@ -21,6 +21,7 @@
 
 import { WebError } from '@deepseek-ai/dsh-web'
 import type { WebFetchBody, WebFetchProvider, WebFetchRequest, WebFetchResult, WebProviderStatus } from '@deepseek-ai/dsh-web'
+import { clampTimeout, deadline, timeoutOf } from '@deepseek-ai/dsh-timeout'
 import { classifyContentType, decoderForCharset, isSameOrigin, parseCharset, validateFetchUrl } from './policy.ts'
 
 /** Resolved provider limits (the plugin's schemastery Config supplies defaults). */
@@ -56,35 +57,25 @@ export class LocalFetchProvider implements WebFetchProvider {
   }
 
   async fetch(request: WebFetchRequest, exec?: { readonly signal?: AbortSignal }): Promise<WebFetchResult> {
-    const timeoutMs = request.timeoutMs !== undefined
-      ? Math.min(request.timeoutMs, this.limits.maxTimeoutMs)
-      : this.limits.timeoutMs
+    if (exec?.signal?.aborted) throw new WebError('web fetch aborted', 'WEB_ABORTED')
+    const timeoutMs = clampTimeout(request.timeoutMs, this.limits.timeoutMs, this.limits.maxTimeoutMs)
 
-    // One controller drives both the caller's abort and our own timeout, so the
-    // network request and the streaming read both stop on either.
-    const controller = new AbortController()
-    const onAbort = (): void => { controller.abort() }
-    if (exec?.signal !== undefined) {
-      if (exec.signal.aborted) throw new WebError('web fetch aborted', 'WEB_ABORTED')
-      exec.signal.addEventListener('abort', onAbort, { once: true })
-    }
-    const timer = setTimeout(() => { controller.abort(new WebError('web fetch timed out', 'WEB_FETCH_TIMEOUT')) }, timeoutMs)
-
-    try {
-      return await this.followAndRead(request.url, controller)
-    } finally {
-      clearTimeout(timer)
-      if (exec?.signal !== undefined) exec.signal.removeEventListener('abort', onAbort)
-    }
+    // One deadline signal fuses the caller's abort with our own timeout, so the
+    // network request and the streaming read both stop on either. The timeout
+    // abort carries a TimeoutReason we recover afterward to classify the cause
+    // (translateAbortOrNetwork), instead of hand-rolling a controller + timer +
+    // reason-recovery dance.
+    using d = deadline(exec?.signal, timeoutMs, 'WEB_FETCH_TIMEOUT')
+    return await this.followAndRead(request.url, d.signal)
   }
 
   /** Follow same-origin redirects up to the hop cap, then read the final response. */
-  private async followAndRead(initialUrl: string, controller: AbortController): Promise<WebFetchResult> {
+  private async followAndRead(initialUrl: string, signal: AbortSignal): Promise<WebFetchResult> {
     let currentUrl = validateFetchUrl(initialUrl, this.limits.maxUrlLength)
     let redirectsFollowed = 0
 
     for (;;) {
-      const response = await this.requestOnce(currentUrl, controller)
+      const response = await this.requestOnce(currentUrl, signal)
 
       if (isRedirectStatus(response.status)) {
         // The redirect budget is enforced BEFORE this hop's target is resolved
@@ -127,20 +118,20 @@ export class LocalFetchProvider implements WebFetchProvider {
         continue
       }
 
-      return await this.readBody(response, currentUrl, controller.signal)
+      return await this.readBody(response, currentUrl, signal)
     }
   }
 
-  private async requestOnce(url: URL, controller: AbortController): Promise<Response> {
+  private async requestOnce(url: URL, signal: AbortSignal): Promise<Response> {
     try {
       return await fetch(url, {
         method: 'GET',
         redirect: 'manual',
         headers: { 'user-agent': this.limits.userAgent, 'accept': 'text/html,application/xhtml+xml,text/*;q=0.9,application/json;q=0.8' },
-        signal: controller.signal,
+        signal,
       })
     } catch (error: unknown) {
-      throw translateAbortOrNetwork(error, controller.signal)
+      throw translateAbortOrNetwork(error, signal)
     }
   }
 
@@ -255,24 +246,18 @@ function resolveRedirect(location: string, base: URL): URL {
 }
 
 /**
- * Translate a thrown fetch/stream error into a `WebError`. Our own
- * `WEB_FETCH_TIMEOUT` (passed to `controller.abort(reason)`) and any other
- * already-typed `WebError` pass through; an `AbortError` becomes `WEB_ABORTED`,
- * UNLESS the abort was our timeout — the body-read reader surfaces a generic
- * `AbortError` rather than the abort reason, so we recover the timeout's
- * `WebError` from `signal.reason`; anything else is a transport/network failure
- * (`WEB_PROVIDER_ERROR`).
+ * Translate a thrown fetch/stream error into a `WebError`, classified by the
+ * deadline signal rather than the error's shape (which differs by phase: the
+ * request-phase `fetch` rejects with the abort reason, while the read-phase
+ * reader surfaces a bare `AbortError`). `timeoutOf(signal, 'WEB_FETCH_TIMEOUT')`
+ * recovering OUR reason means our timeout fired (`WEB_FETCH_TIMEOUT`); any other
+ * abort — an upstream cancel, or a foreign/outer deadline's timeout under
+ * nesting — is `WEB_ABORTED`; a throw with the signal NOT aborted is a
+ * transport/network failure (`WEB_PROVIDER_ERROR`).
  */
-function translateAbortOrNetwork(error: unknown, signal?: AbortSignal): WebError {
-  if (error instanceof WebError) return error
-  if (error instanceof DOMException && error.name === 'AbortError') {
-    // A timeout abort carries its WebError as the signal reason; honor the
-    // WEB_FETCH_TIMEOUT contract instead of reporting a generic cancellation.
-    // (Node rejects WITH the reason — the WebError branch above — so this only
-    // fires on a runtime that surfaces a bare AbortError while reason is set.)
-    /* v8 ignore next */
-    if (signal?.reason instanceof WebError) return signal.reason
-    return new WebError('web fetch aborted', 'WEB_ABORTED', { cause: error })
-  }
+function translateAbortOrNetwork(error: unknown, signal: AbortSignal): WebError {
+  const timeout = timeoutOf(signal, 'WEB_FETCH_TIMEOUT')
+  if (timeout !== undefined) return new WebError('web fetch timed out', 'WEB_FETCH_TIMEOUT', { cause: timeout })
+  if (signal.aborted) return new WebError('web fetch aborted', 'WEB_ABORTED', { cause: error })
   return new WebError(`web fetch failed: ${String(error)}`, 'WEB_PROVIDER_ERROR', { cause: error })
 }

packages/web/web-fetch-local/tsconfig.json

@@ -17,6 +17,9 @@
     {
       "path": "../../../vendor/schemastery"
     },
+    {
+      "path": "../../util/timeout"
+    },
     {
       "path": "../web"
     }