The sandbox execute wrapper JSON round-tripped the return and blindly cast it
to ToolExecuteReturn. A JSON-valid but wrong-shape return — a bare string,
{ content: 'ok' }, blocks without a type tag — sailed through: the registry
spreads result.content, so { content: 'ok' } became ['o','k'], passed the
session log's isJsonValue gate, and the DeepSeek serializer then flattened it
to '(no output)' — silent corruption of the next model request and every
replay, instead of a contained tool error.
The round-tripped value is now shape-checked against the two ToolExecuteReturn
forms (array of content blocks, or { content: blocks, meta? }); block checks
are structural only (plain object + string type tag) because the ContentBlock
union is merge-extensible. A wrong shape — and the formerly cryptic
forgot-return/bare-string cases — fails that one call with a teaching error
echoing a truncated preview of what was returned and the two valid forms.
New specs pin the object-form pass-through (meta included), six rejection
shapes, and the preview truncation; per-file 100% coverage holds.
P1 review finding: extractMeta timed only the literal's vm evaluation;
materializing the RESULT then read properties ordinarily on the HOST
stack, so a meta literal smuggling a getter (get name() { while(true){} })
could wedge the host outside any timeout — defeating the exact spin
isolation the worker thread exists for.
Rather than harden the evaluator (descriptor walks, AST validation),
delete the mechanism: the workflow's identity now reaches the seam as a
plain JSON field (WorkflowStartRequest.meta), carried by the tool as a
schema-validated `meta` object parameter the model fills directly. The
engine only shape-validates data (validateMeta, every violation named)
and pre-parses the body; the scanner, the vm evaluation, and the
host-side materialization are gone, and with them the hole. A body
still opening with a Claude Code-style `export const meta` statement
gets a pointed SCRIPT_PARSE message (the likeliest authoring slip; a
CC script's body stays drop-in, only its meta header moves into the
parameter). syncTimeoutMs now governs exactly one thing: the initial
synchronous slice inside the worker.
The RFC's decision section is rewritten in place (implemented-RFC
rule); the embedded-meta format moves to alternatives-considered with
the hole as the reason. Tool description, presentation (title now reads
meta.name directly — the textual sniff is gone), seam vocabulary docs,
and catalogs follow.
Pure mechanical rename now that the package's internals are the
worker-thread engine: directory, package name, spec/e2e filenames,
module tags and logger prefixes, tsconfig/knip/run-gates/AGENTS.md
references, example cordis.yml plugin ids, doc links; catalogs
regenerated and the lockfile refreshed.
The outer ring catches up with the engine swap (the package's own
README/JSDoc rode the port commit):
- Seam module doc and README name the worker-thread engine as THE
implementation, with isolated-vm/separate-process sandboxing as the
deferred hardening; the seam service doc states the holder-owned-runs
contract (engine-fiber disposal deliberately leaves live runs to
their holders).
- Seam contract precision: agentsStarted documents the termination-path
degradation to the host-observed count; the events section scopes the
agent-start/agent-end pair to calls that STARTED a child run;
WorkflowRun wording drops the vm-era abandonment language.
- The dynamic-workflows RFC is rewritten in place to the shipped
mechanism (implemented-RFC rule): why worker threads, the thread's
concrete buys, the in-process node:vm first cut recorded under
alternatives considered; the tool section describes the usage policy
as the tool's own prompt section.
- gen-doc-graphs: six workflow/* DYNAMIC_EVENT_DISPATCHERS entries (the
catalog no longer claims nothing dispatches them) and the seam-note
wording; core-data-structures gains its workflow.md index row;
packages/README + AGENTS.md layout line + example cordis.yml comments
say worker-thread; catalogs regenerated.
The sandbox docs overclaimed a containment contract the design never makes:
"capability access is routed through cordis services, never Node built-ins,
so everything a mounted plugin does stays inspectable and disposable". The
host-realm helpers on the sandbox global (harness, console, btoa) are
reachable functions, so mount code that goes looking can reach the host realm
through one of them — accepted under the trust stance, because the ctx a
mount ultimately receives is fully privileged anyway. Reword the sandbox
module doc, the README trust stance, and the RFC sandbox-semantics section to
say exactly that: the traps and small global surface STEER honest code onto
the cordis services; they are not a security boundary.
Two review findings (#220) on the sandbox context façade:
- Undeclared services were reachable: the façade resolved any live global via
ctx.get(name), so ctx.bash worked without inject: ['bash']. A cross-mount
consumer could then depend on a provider cordis never saw — unmounting the
provider would neither park the consumer nor unwind its registered tools,
leaving a model-visible tool that fails only at execution. The façade now
reads ctx.fiber.inject and refuses any service the mount did not declare
(with a teaching error naming the inject fix), so the dependency is always
visible to cordis and its activation/unload semantics bind.
- ctx.tools.get returned the live ToolDefinition, including execute — mount
code could call another tool directly and bypass ToolRegistry.execute and
its pre/post-execute hooks and accounting. get now returns the same
read-only name/description/parameters view as schemas(), never an invocable.
Adds inject-gate and schema-view regression cases to sandbox-context.spec.ts
(undeclared property/get denied, declared allowed, the cross-mount zombie-tool
scenario refused at call time, get exposes no execute). Package stays at
per-file 100% coverage. RFC, mount description, and tool-catalog updated.
Review finding (#220): the guarded proxy only special-cased ctx.tools, so
mount code could reach an UNGUARDED context through ctx.root, ctx.extend(), or
a service instance's .ctx, then ctx.root.tools.register({…}) to bypass the
marker check and host-realm normalization — a raw vm-realm result would later
error a real agent turn at the session-log plainness check.
The sandbox ctx is now a whitelist façade, not a pass-through proxy: it exposes
only what a mount needs — tools.register (marker-guarded), on/once, provide, the
timer helpers, and injected services resolved through a guarded get — and denies
every framework-plumbing member (root, parent, fiber, reflect, registry, extend,
isolate, intercept, plugin, set, mixin, …) with a teaching error. Injected
services are wrapped so a method returning a Context is rejected on the way back
(the .ctx escape), closing the one indirect leak. There is no context-valued
member left to reach; cross-mount provide/inject is untouched (the plugin's own
inject and the fiber's pending/active gating are unchanged). ctx.plugin (child
plugins) and ctx.set are denied by design; ctx.effect is deferred (FIXME).
Adds tests/sandbox-context.spec.ts covering the escape class (root/extend/fiber/
plugin/set/… denied, the classic root.tools.register bypass, the .ctx escape,
read-only writes) plus the async-service and symbol/in-operator paths for 100%
coverage. RFC/README/tool-catalog/config-catalog updated; api-catalog.ts
regenerated (also picks up the codeRuntime service that entered on the master
merge and was left stale).
The tree SHAPE was the wrong surface for the model: what it needs from
cordis_inspect is what services, plugins, and capabilities are loaded, not the
fiber hierarchy. The plugins section is now a flat name + lifecycle-state list
from ctx.registry (deterministically sorted, one line per instance); the ASCII
tree renderer, the parent→child rebuild, and the dyn-id tree annotation are
deleted — dynamic mounts keep their own richer dynamic section (id, state,
provides, waits). Net -49 lines; RFC and READMEs state the flat-list contract.
Field sessions showed models writing tool schemas in the JSON-Schema dialect
by strong prior — type: 'integer', required: false, then the full
{ type:'object', properties, required: [...] } wrapper — and the rejection
text itself pushed a nearly-correct DSL attempt BACK to raw JSON Schema: one
stats tool cost three consecutive schema errors before mounting. The boundary
now normalizes wherever the input has exactly one meaning (wrapper unwrapped
with the required array becoming per-property flags at any nesting level,
integer → number, required: false → optional, all rebuilt as fresh host-realm
objects) and rejects only genuinely meaningless input, enumerating the valid
vocabulary in the error. Re-running the failing session mounts first-try.
The mount description documents both accepted forms.
The design record for tool-cordis: the three-tool contract, the vm sandbox
trust stance and boundary mechanisms, the dynamic-group lifecycle, cross-mount
provide/inject composition, the generated runtime API catalog, and the
alternatives weighed (per-capability registration tools, hand-maintained API
tables, a mount provenance event, a hardened sandbox).
The seam shipped without its own RFC — the reconstructable-requests RFC
was amended with the mechanics, but the decision record (why a
compose-once frozen prefix, and what the per-request before/after shape,
a system-prompt section, a durable history opener, per-turn composition,
and a dedicated session event each lost to) had no home. Implemented
lifecycle, feature class, dated to the first commit of the work.
The Date.now()/Math.random()/argless-new-Date() bans existed solely to
keep scripts resume-compatible for the deferred journaling/resume
feature. Pre-support for an unimplemented feature is speculative cost:
scripts may now read the clock freely; implementing resume reintroduces
the bans as a script-contract tightening. The RFC's shipped-state
description is updated in place, the tool DESCRIPTION drops the
constraint sentence (the pinned text-turn header follows), and the
engine README's trust-premise paragraph now leans on absent globals
alone.
Review follow-up (#196): a listed name with no registered tool was silently
ignored; misconfiguration must block work instead. The check lives in the
assembly — the earliest moment the registered tool set exists (tool plugins
register after the service constructs) and the only universal one (cordis has
no "all plugins loaded" event; registrations change at any time). assemble()
is now async so the throw surfaces as a rejection rather than a synchronous
escape from a Promise-returning method.
Blast radius, pinned by a loop-level test: the rejection reaches the turn's
outer catch — the turn closes balanced with an `error` reason, agent/error
mirrors it, no step opens, no request/header is logged, no request reaches
the adapter, and the agent returns to idle; every turn fails identically
until the config is fixed. A boot-time validation pass was considered and
rejected (recorded in the RFC). The general principle — misconfiguration
fails loud, never a silent skip — is added to AGENTS.md.
A three-dot rest entry reads as elision in a cordis.yml; the spelled-out
sentinel says what lands there. The literal now appears once in code (the
constant) and once in the value-pinning test; every other reference — the
forwarding tests included — imports TOOL_ORDER_REST, which adds the
dsh-system-prompt devDependency to the two app packages.
Review follow-up on #196.
The #192 review discussion, preserved where design rationale lives: a
schema-object library (zod/schemastery) cannot sit at a wire-data boundary;
ajv replaces only the value walker while the subset gate — the module's
point — stays hand-written; provider JSON mode guarantees valid JSON, not
schema-conforming JSON, and would trade away mid-run tools and in-turn
validation retry for it. Strict tool schemas are named as the accepted
upgrade path when the provider ships them.
The parent branch dropped structuredNudgeRetries; the two workflow-side test
setups stop passing it, and the RFC's foundation paragraph now describes the
current design (final-ASSEMBLY enforcement logged via request/header, the
post-capture pre-execute deny, the start() schema snapshot, no re-prompt)
instead of the retired agent/request + nudge shape.
The review's critical finding: the engine README and the tool
description read as if the missing filesystem/network/Node globals were
enforced, but a script can reach the host Function constructor via
globalThis.constructor.constructor and from it process and every Node
builtin. Per the trust premise this is ACCEPTED (model-written scripts,
bash-equivalent trust; genuine sandboxing is the deferred engine swap
already listed) — but the docs must say so instead of implying a wall.
The trust-premise sections (engine README, module doc, RFC) now name
the escape and its acceptance; the model-facing tool description says
the APIs are not PROVIDED rather than implying they are prevented.
Beyond the mechanical conflicts (provider capability lines vs master's new
inheritsParentContext field; generated catalogs regenerated rather than
hand-merged; knip/lockfile), three master-side reworks required semantic
adaptation of this branch:
- The persona rework removed AgentOptions.systemPrompt, which was the
structured-output instruction's channel. The instruction now rides the
SAME final-request enforcement listener that injects the schema'd tool:
appended per request to final.system (per-request wire state, not agent
prompt state). Tests assert the wire request (adapter.requests) instead
of child.options; the bare-direct-dispatch test pins the no-system arm.
- Tool guidance moved out of deployment prompts into per-tool prompt
sections; the examples' workflow paragraph became a tool:<toolName>
section contributed by dsh-tool-workflow (explicit-ask-only policy),
and both example personas resolve to master's minimal identity+behavior
form. tool-workflow gains inject: systemPrompt (+ peer dep, tsconfig
ref); the export-shape guard updated.
- The uniform-RFC-format gate: the dynamic-workflows RFC restructured to
the implemented/ skeleton (bare Status line; Proposal -> Decision;
What-was-rejected -> Alternatives considered; new Consequences), and
the overall-run-timeout deferral is now recorded in the RFC's Deferred
list. The doc-graphs atlas classification gains the workflows seam
(workflow-vm implementation, tool-workflow consumer).
Master's harness-identity section made "empty assembled prompt" states
unreachable through the loop, so the instruction-append is a plain
undefined-ternary and the structured tests assert append-not-replace.
All snapshot goldens (including workflow-run) replay unchanged. Full
local CI-equivalent gate sequence green on the merged tree.
Codex convergence round 1 on the review-response commits: agent()'s
post-acquire window was real and unguarded. `await acquireSlot()` yields
at least one microtask tick even when a slot is free (and a queued
waiter resumes a tick after its release), so a cancel() landing in that
tick let the continuation start a child carrying an ALREADY-aborted
signal — the in-code comment claimed the window could not exist. A
provider that subscribes only to future abort events (the test stub;
the seam does not promise pre-aborted-signal handling) would never
settle such a child, leaking it until the dispose grace abandoned the
run, and a backend that misses the pre-aborted signal would burn a full
model turn after the user cancelled.
agent() now re-checks isCancelled() immediately after the acquire
(inside the slot-owning try, so the finally still releases), making
every post-cancel path reject before subagents.start. New deterministic
regression: cancel() in the same synchronous frame as start() lands in
the free-slot await tick — the run settles cancelled with ZERO children
started (previously: one leaked child and a grace-delayed settle). The
raced-release test's comment now states what it actually pins (the
queued-waiter rejection path). Also aligns the RFC's auto-concurrency
formula with the code (min(16, max(1, availableParallelism() - 2))).
Two review responses that belong together — the same review argued the
engine was defending the wrong threat while a benign-input bug wedged
the product.
1) Drop hostile-value containment; state the trust premise.
Scripts are model-written — the same trust level as the model's bash
access — yet successive pre-push review rounds had ratcheted in defenses
that only matter against an adversarial author: trap-free proxy
rejection, accessor-never-invoked descriptor walks, realm-side
pre-rendering of thrown values, realm-built promises/arrays/error clones
with structural fatal recognition. That same author keeps a documented,
accepted, unkillable event-loop spin, so containing its error VALUES is
cost without a threat model — and the planned hardened engine
(worker/isolated-vm) gets value isolation by serialization and deletes
all of this machinery anyway.
What stays, because benign scripts hit it constantly: result never
rejects; dropped hook promises cannot become unhandled rejections; the
value boundary rejects LOUD everything JSON cannot carry (now a plain
recursive walk — getters are read ordinarily and their result is what
crosses; a throwing read fails loud); a "__proto__" key still copies as
a data property; the fatal-vs-null combinator discipline (now host
instanceof — unforgeable from the realm and simpler than clone-shape
recognition). What changes for scripts (documented in the engine
README): hooks hand back host values and host errors — in-script
`instanceof Error` on a hook failure is false (branch on e.name/e.code)
— and args are host-cloned once so a script cannot mutate the caller's
object. realm.ts drops 289 → 173 lines; the hostile-value test tables go
with it. The premise now leads the engine module doc, the README, and
the RFC's engine section, with the removed machinery recorded under
What was rejected.
2) result settles within the dispose grace of a cancellation.
Review finding (verified through the real registry + tool + engine): a
script parked on a promise no hook owns — `await new Promise(() => {})`,
`await Promise.race([])`, a returned never-settling thenable — could not
be settled by cancel(): hooks reject and children abort, but nothing
touches a promise the engine does not own, so `result` stayed pending
FOREVER (the previous cut even pinned that as intended). The tool awaits
run.result BEFORE its disposing finally, the registry awaits the tool,
the loop awaits the registry — one such script wedged the whole agent
turn past any abort, unrecoverable in-process; the mock engine in the
tool's abort test settles result on cancel, which is exactly the
behavior the real engine lacked, so no existing test could see it.
The seam contract now says it out loud: once a run is cancelled, result
SETTLES within the implementation's bounded grace even if the script
never does. The vm engine arms an abandon channel in cancel(); drive()
races the script against it, force-settling 'cancelled' at the grace
(the abandoned settlement stays contained; a post-slice synchronous spin
remains the documented limitation). dispose()'s outer race now exists
for child quiescence only, and `workflow/end` again fires exactly once
per started run. The old 'result stays pending' pin is FLIPPED to the
new contract (the pinned behavior was the bug); new regressions cover
cancel-then-settle on a parked script, a never-settling returned
thenable, and the full composition through the REAL registry + tool +
vm engine (tool-workflow gains workflow-vm/subagent devDeps for it).
agentsStarted JSDoc clarified while touching the vocabulary (accepted
calls, including ones still queued at cancellation).
Define the in-file RFC contract in docs/rfc/README.md § The file format:
the header block (`# RFC: <title>` plus a dateless Status enum
cross-checked against the lifecycle folder), the per-lifecycle body
skeleton (a Problem opener everywhere; Proposal/Alternatives considered/
Acceptance criteria/Risks in proposed/; present-tense Decision/
Consequences with proposal-era headings banned in implemented/; the
frozen proposal shape in rejected/), and a mandatory Alternatives
considered section with a date-fenced grandfather comment for pre-format
RFCs whose alternatives are not reconstructible from the record.
Enforce it with a new doc-sync gate, scripts/verify-rfc-format.ts, and
normalize all 112 RFCs to it: ~15 Status-line spellings collapse to the
enum, 29 Context openers become Problem, the 39 legacy-format XXX debt
markers are resolved and banned from reappearing, proposal-era sections
in implemented RFCs are rewritten to shipped reality (including the
web/fs/subagent seam RFCs' migration plans and test checklists, closing
the doc-tiers deferred-work item on the web seam), every RFC gains an
Alternatives considered section or the grandfather comment, and the
bilingual pair is re-mirrored and re-recorded.
Move the generated index tables out of README.md into a fully generated
docs/rfc/INDEX.md — gen-rfc-index now writes the whole file, and
verify-rfc-classification checks its freshness and rejects index-shaped
rows in the curated README — which makes room for the format contract to
live in the README front door instead of a separate FORMAT.md.
The decision record, and the first RFC written in the new format, is
docs/rfc/implemented/process/2026-07-05-uniform-rfc-format.md.
Codex code-review round 5: agent()/parallel()/pipeline() returned HOST Promise
objects into the script realm — Object.getPrototypeOf(agent('x')) reached host
Promise.prototype, contradicting the realm contract (correctness containment,
not the accepted sandbox stance). The rejection channel had the same leak one
hop away: a caught hook failure was a host WorkflowError (host Error.prototype
chain), and phase()/log() threw host errors synchronously.
All three surfaces are realm-built now:
- hook promises: the realm's own Promise.resolve (bound at context setup)
assimilates the host promise, so the script-visible promise carries realm
prototypes; the realm promise gets the same no-op rejection consumer as the
host one (a script may drop it).
- hook failures: rejections and phase/log sync throws are translated at the
boundary into realm-built clones (name/code/message/fatal via an in-realm
factory); non-WorkflowError host failures become generic realm Errors
carrying their describeThrown rendering.
- the combinators recognize FATAL clones structurally
(isFatalWorkflowErrorClone: proxy-guarded descriptor reads), preserving the
fatal-vs-null discipline across the boundary; a script forging the shape
kills only its own run. drive() maps any post-cancel failure to 'cancelled'
by run state (a CANCELLED clone deliberately fails the host instanceof).
Tests: realm-promise identity for all three hooks + host Promise.prototype
pollution unreachable; clone shape (instanceof realm Error, name/code/fatal/
message) with prototype-chain mutation staying realm-side; a rejecting
provider result crossing as a generic clone; phase/log sync-throw clones;
combinator catch branches (string throw, proxy throw, shape-miss forgery →
null; forged fatal → kills own run); existing fatal-propagation, cancellation,
and unhandled-rejection tests as canaries.
Codex code-review round 4 flagged the return channel: an async IIFE
Promise-assimilates a returned thenable, so its then() runs past the sync
slice and the RESOLUTION replaces the raw object. Verified against the real
engine and judged behavior, not defect:
- Assimilation is standard JavaScript (an async function's returned thenable
resolves before the caller sees it) and is load-bearing ergonomics: an
un-awaited 'return agent(...)' / 'return parallel(...)' resolves to the
intended value precisely because of it. Rejecting callable-then returns
would break that; intercepting pre-assimilation is spec-impossible (the
Get(v,'then') and job enqueue are internal to promise resolution).
- The realm-boundary guard applies to the RESOLUTION (a thenable resolving to
non-JSON is still RESULT_UNSERIALIZABLE), so nothing crosses unmaterialized.
- A spin inside a returned thenable's then() is the same accepted class as any
post-slice spin (it runs on the microtask queue, past the vm timeout's
reach); the docs previously said 'after the first await', which was too
narrow — reworded to 'past the initial synchronous slice (an await
continuation, or a thenable's then invoked by promise resolution)'.
Pinned with an engine test (un-awaited return agent(); custom thenable
resolution as the return value; thenable resolving to non-JSON rejects), and
the limitation wording updated in the module doc, README, and RFC.
Codex code-review round 3: the round-2 'contained stack getter' still let a
script escape the vm sync-slice timeout — throw { get stack() { while(true){} } }
put the spin on the HOST catch path, where no timeout applies (verified: a
direct sync-slice spin dies by the timeout; the getter-hidden one hung the
process). Identity-trusting the native getter is also insufficient: V8 stack
formatting reads script-controllable hooks at format time (Error.prepareStackTrace,
a subclass name getter — both empirically confirmed), so ANY host-side
formatting of a realm error can run realm code.
The fix moves rendering into the realm itself: the compiled body (and the meta
literal) is wrapped in a realm-side catch that pre-renders the thrown value to
a string (REALM_THROWN_RENDERER_SOURCE) — a hostile accessor/toString now runs
as ordinary script code, killed by the sync-slice timeout or falling under the
documented post-await spin limitation; host WorkflowErrors pass through for
the CANCELLED mapping. The host catch descriptor-reads the pre-rendered string
(thrownRendering) or falls back to describeThrown, which invokes no getter
whose identity is not the host realm's own native stack getter.
Tests: hostile-table expectations updated for realm-side rendering; new
regressions for the getter-hidden sync spin dying by the vm timeout (engine +
meta paths) and for a hostile thenable rejection that bypasses the realm
wrapper (renders host-side, proxy labelled, traps never run); describeThrown/
thrownRendering unit tables including the realm-error identity-mismatch case.
