fix(sandbox): preserve filesystem cwd semantics

This commit is contained in:
Tianyi Cui
2026-07-21 19:53:02 +08:00
parent 7677d49c93
commit ae98294ec5
14 changed files with 119 additions and 27 deletions

View File

@@ -17,12 +17,12 @@ The plugin also contributes the `tool:bash` prompt section (order 105): check th
| `command` | string (required) | Run via `bash -c`. No state persists between calls — use `workdir`, not `cd`. |
| `description` | string (required) | One-line, active-voice summary of the command (5-10 words), for UI/log display only — no effect on execution. |
| `timeoutMs` | number | Timeout override in milliseconds. The executor applies its configured default and cap. |
| `workdir` | string | Working directory for this call. Defaults to the calling agent's session cwd (`session.header.cwd`) so each session runs in its own workspace; a relative `workdir` is resolved against that session cwd. |
| `workdir` | string | Working directory for this call. Defaults to the filesystem identity of the calling agent's session cwd (`session.header.cwd`) so each session runs in its own workspace; a relative `workdir` is resolved against that same identity. |
| `run_in_background` | boolean | Return a task id immediately; no timeout applies. |
| `sandbox_permissions` | string enum | ADVERTISED ONLY when the mounted executor sandboxes (`ctx.bash.sandboxMode` reports a confining default): the wider mode a denied command needs, from the closed target vocabulary `workspace-write`/`danger-full-access` (never cut down to the executor's default — the effective mode is per-session; strict widening is checked at execution against it, and a non-widening request fails without prompting anyone). |
| `justification` | string | Required together with `sandbox_permissions` (each without the other is a validation error): one sentence for the user explaining why this exact command needs the wider access. |
`command`, `workdir`, and `timeoutMs` are resolved against the executor's config defaults via `ctx.bash.resolve()` before execution, so the executor seam (`BashExecSpec`) receives explicit `workdir`/`timeoutMs` values. The workdir default is applied in the tool layer (from the calling agent's `session.header.cwd`) BEFORE `resolve()` — the per-session cwd must come from `exec.agent`, since N sessions share one executor; only when no session cwd is available does the executor fall back to its own config / `process.cwd()`.
`command`, `workdir`, and `timeoutMs` are resolved against the executor's config defaults via `ctx.bash.resolve()` before execution, so the executor seam (`BashExecSpec`) receives explicit `workdir`/`timeoutMs` values. The workdir default is applied in the tool layer from the calling agent's `session.header.cwd` BEFORE `resolve()` — the per-session cwd must come from `exec.agent`, since N sessions share one executor; only when no session cwd is available does the executor fall back to its own config / `process.cwd()`. When sandbox policy is present, the tool reuses its already-canonical `workspaceRoot` as the workdir base so confinement and process launch cannot resolve the same session spelling differently.
### Managed shell environment

View File

@@ -19,7 +19,7 @@ import type {} from '@deepseek-ai/dsh-system-prompt'
import type {} from '@deepseek-ai/dsh-tasks'
import type {} from '@deepseek-ai/dsh-user-approval'
import type { SandboxExecutionPolicy, SandboxMode } from '@deepseek-ai/dsh-sandbox'
import { ESCALATION_TARGETS, approveEscalation, validateEscalationArgs } from '@deepseek-ai/dsh-sandbox'
import { ESCALATION_TARGETS, approveEscalation, canonicalPath, validateEscalationArgs } from '@deepseek-ai/dsh-sandbox'
import type { SandboxPolicyService } from '@deepseek-ai/dsh-sandbox-policy'
import { DSH_ENV_PREFIX } from '@deepseek-ai/dsh-bash'
import type { DshEnvironment, DshEnvironmentKey } from '@deepseek-ai/dsh-bash'
@@ -299,11 +299,18 @@ function presentBashResult(args: unknown, result: ToolResult): ToolResultView |
}
/**
* Resolve an explicit workdir first, making a relative one session-cwd-relative;
* otherwise use the session cwd and leave executor defaulting as the fallback.
* Resolve an explicit workdir first, making a relative one session-workspace-relative;
* otherwise use the filesystem identity of the session cwd and leave executor
* defaulting as the fallback. A resolved sandbox-policy root wins so workdir
* and confinement use the exact same per-call identity.
*/
function resolveWorkdir(modelWorkdir: string | undefined, exec: { agent?: Agent }): string | undefined {
const sessionCwd = exec.agent?.session.header.cwd
function resolveWorkdir(
modelWorkdir: string | undefined,
exec: { agent?: Agent },
policyWorkspaceRoot?: string,
): string | undefined {
const headerCwd = exec.agent?.session.header.cwd
const sessionCwd = policyWorkspaceRoot ?? (headerCwd === undefined ? undefined : canonicalPath(headerCwd))
if (modelWorkdir === undefined) return sessionCwd
if (sessionCwd !== undefined && !isAbsolute(modelWorkdir)) {
return resolvePath(sessionCwd, modelWorkdir)
@@ -418,7 +425,7 @@ export function apply(ctx: Context, config: Config = {}): void {
const policy = approvedMode === undefined
? standingPolicy
: { ...(standingPolicy as SandboxExecutionPolicy), mode: approvedMode }
const workdir = resolveWorkdir(args.workdir, exec)
const workdir = resolveWorkdir(args.workdir, exec, standingPolicy?.workspaceRoot)
const dshEnv = bashEnv.collect(exec)
const request = {
command: args.command,

View File

@@ -1,5 +1,5 @@
import { spawnSync } from 'node:child_process'
import { mkdtemp, readFile, rm } from 'node:fs/promises'
import { mkdir, mkdtemp, readFile, rm, symlink } from 'node:fs/promises'
import { homedir } from 'node:os'
import { basename, join } from 'node:path'
import { afterEach, beforeEach, describe, expect, it } from 'vitest'
@@ -146,4 +146,50 @@ describe('one-context multi-project sandbox', () => {
await expectMissing(join(projectB, 'from-a.txt'))
await expectMissing(join(projectA, 'from-b.txt'))
})
it.skipIf(!processSandboxUsable)('keeps symlink-sensitive session cwd semantics aligned across bash, fs, and policy', async () => {
const active = ctx as Context
const lexicalRoot = await projectDir('lexical-workspace')
const physicalRoot = await projectDir('physical-workspace')
const physicalChild = join(physicalRoot, 'child')
await mkdir(physicalChild)
const link = join(lexicalRoot, 'link')
await symlink(physicalChild, link, process.platform === 'win32' ? 'junction' : 'dir')
const sessionCwd = `${link}/..`
const handle = await active.agents.create({
sessionId: SessionId('symlink-parent-session'),
meta: { cwd: sessionCwd },
})
const [bashOwn, bashLexical, fsOwn, fsLexical] = await Promise.all([
active.tools.execute({
callId: CallId('bash-symlink-own'), name: 'bash', agent: handle.agent,
arguments: { command: 'printf bash > bash-owned.txt', description: 'Write physical workspace marker' },
}),
active.tools.execute({
callId: CallId('bash-symlink-lexical'), name: 'bash', agent: handle.agent,
arguments: { command: `printf escaped > ${join(lexicalRoot, 'bash-escaped.txt')}`, description: 'Attempt lexical workspace write' },
}),
active.tools.execute({
callId: CallId('fs-symlink-own'), name: 'write', agent: handle.agent,
arguments: { file_path: 'fs-owned.txt', content: 'fs' },
}),
active.tools.execute({
callId: CallId('fs-symlink-lexical'), name: 'write', agent: handle.agent,
arguments: { file_path: join(lexicalRoot, 'fs-escaped.txt'), content: 'escaped' },
}),
])
expect(bashOwn.isError).toBe(false)
expect(resultText(bashOwn)).not.toContain('[sandbox:')
expect(bashLexical.isError).toBe(false)
expect(resultText(bashLexical)).toContain('[sandbox: file access denied under workspace-write mode]')
expect(fsOwn.isError).toBe(false)
expect(fsLexical.isError).toBe(true)
expect(resultText(fsLexical)).toContain('[sandbox: file access denied under workspace-write mode]')
expect(await readFile(join(physicalRoot, 'bash-owned.txt'), 'utf8')).toBe('bash')
expect(await readFile(join(physicalRoot, 'fs-owned.txt'), 'utf8')).toBe('fs')
await expectMissing(join(lexicalRoot, 'bash-escaped.txt'))
await expectMissing(join(lexicalRoot, 'fs-escaped.txt'))
})
})

View File

@@ -8,7 +8,7 @@ The filesystem stack: a provider seam (text IO + atomic mutation with an optiona
| `fs-local/` | Local-filesystem `FileSystem` implementation | (registers `ctx.fs`) |
| `fs-sandbox/` | Sandbox-enforcing `FileSystem`: extends `fs-local` and fences write/edit by the per-call mode + workspace root policy (read-only denies, workspace-write contains to the session workspace + temp roots), reads pass through | (registers `ctx.fs`) |
| `fs-policy/` | Policy gate plugin: observed-state + read-before-edit + version-guarded write/edit, via the `fs/*` event gate | (no service — `fs/*` listeners) |
| `tool-fs/` | Model-facing `read`/`write`/`edit` tools AND the executor (reads via `ctx.fs`, owns read windowing, dispatches `fs/*`); advertises the sandbox escalation fields when the mounted `ctx.fs` confines | (registers on `ctx.tools`) |
| `tool-fs/` | Model-facing `read`/`write`/`edit` tools AND the executor (reads via `ctx.fs`, owns read windowing, dispatches `fs/*`); resolves relative paths from the filesystem identity of the session cwd and advertises sandbox escalation fields when the mounted `ctx.fs` confines | (registers on `ctx.tools`) |
| `tool-fs-search/` | Model-facing `glob`/`grep` discovery tools when `rg` is available on the bash executor `PATH`, backed by fixed ripgrep commands through `ctx.bash`, NOT by `ctx.fs` provider methods | (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 — `fs-sandbox` is the first such replacement (an in-process path fence over the shared sandbox mode; see [the cross-family fs sandbox Agent Note](../../.agents/notes/implemented/feature/2026-07-14-cross-family-fs-sandbox.md)). 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. The mode fence and the read-before-edit gate are orthogonal and compose. Discovery (`tool-fs-search/`) deliberately does NOT extend the provider seam: search is a process-backed `rg` workflow on the bash executor, so filesystem backends stay free of a universal search contract; its tools register only when that executor can find `rg`, and its results are follow-up-readable when the bash workdir and the `read` root are the same workspace (the co-located deployment its README documents).

View File

@@ -95,7 +95,7 @@ export function applyEditTool(ctx: Context, sandbox: FsSandboxSurface): void {
// Resolve the per-call sandbox policy (approved mode > session override
// > backend default, plus the session cwd root) BEFORE anything executes.
const sandboxPolicy = await sandbox.resolvePolicy('edit', args, exec)
const target = await ctx.fs.resolve(input.filePath, sessionResolveOptions(exec))
const target = await ctx.fs.resolve(input.filePath, sessionResolveOptions(exec, sandboxPolicy?.workspaceRoot))
// Single-slot decision: the policy plugin returns { version: vObserved } or
// throws FS_NOT_OBSERVED; the bare default is undefined (unconditional edit).
// No stat — the bare default never manufactures a version basis.

View File

@@ -9,6 +9,7 @@
*/
import type { ToolExecution } from '@deepseek-ai/dsh-tools'
import { canonicalPath } from '@deepseek-ai/dsh-sandbox'
/**
* The session workspace cwd for this call, or `undefined` when none applies.
@@ -16,16 +17,18 @@ import type { ToolExecution } from '@deepseek-ai/dsh-tools'
* @returns the calling agent's session cwd, or undefined for a non-agent caller (the backend then applies its own default).
*/
export function sessionCwd(exec: ToolExecution): string | undefined {
return exec.agent?.session.header.cwd
const cwd = exec.agent?.session.header.cwd
return cwd === undefined ? undefined : canonicalPath(cwd)
}
/**
* Resolution options shared by all model-facing filesystem tools.
* @param exec - the tool-execution context supplying session cwd and cancellation.
* @param policyWorkspaceRoot - resolved per-call root, when a mutation carries sandbox policy.
* @returns provider resolution options for the current tool call.
*/
export function sessionResolveOptions(exec: ToolExecution): { cwd?: string; signal?: AbortSignal } {
const cwd = sessionCwd(exec)
export function sessionResolveOptions(exec: ToolExecution, policyWorkspaceRoot?: string): { cwd?: string; signal?: AbortSignal } {
const cwd = policyWorkspaceRoot ?? sessionCwd(exec)
return {
...cwd !== undefined ? { cwd } : {},
...exec.signal !== undefined ? { signal: exec.signal } : {},

View File

@@ -80,7 +80,7 @@ export function applyWriteTool(ctx: Context, sandbox: FsSandboxSurface): void {
// > backend default, plus the session cwd root) BEFORE anything executes;
// an escalating call throws its distinct text on any non-grant.
const sandboxPolicy = await sandbox.resolvePolicy('write', args, exec)
const target = await ctx.fs.resolve(input.filePath, sessionResolveOptions(exec))
const target = await ctx.fs.resolve(input.filePath, sessionResolveOptions(exec, sandboxPolicy?.workspaceRoot))
// Single-slot decision: the policy plugin produces createIfAbsent/
// replaceIfVersion; the bare default is undefined (unconditional). No stat.
const intent = await ctx.waterfall('fs/write-intent', target, exec, () => undefined)

View File

@@ -9,11 +9,11 @@ Two families enforce the same mode vocabulary: the sandboxed bash executor (`@de
## Config
- `mode` — the deployment default `SandboxMode` (`read-only` / `workspace-write` / `danger-full-access`), validated at load. Default `read-only` (fail-safe).
- `workspaceRoot` — the fallback directory `workspace-write` may write under for agentless calls or sessions without a cwd. Default `process.cwd()`, resolved absolute either way. A normal agent call uses its session header's immutable `cwd` instead.
- `workspaceRoot` — the fallback directory `workspace-write` may write under for agentless calls or sessions without a cwd. Default `process.cwd()`, resolved to its absolute filesystem identity either way. A normal agent call uses its session header's immutable `cwd` instead.
## Surface
- `ctx.sandboxPolicy.resolve({ session?, mode? })` — resolves one complete per-call policy. An explicit approved mode outranks the session's last `sandbox/mode` event, which outranks `defaultMode`; the session's immutable `cwd` becomes `workspaceRoot`, otherwise the configured fallback applies.
- `ctx.sandboxPolicy.resolve({ session?, mode? })` — resolves one complete per-call policy. An explicit approved mode outranks the session's last `sandbox/mode` event, which outranks `defaultMode`; the session's immutable `cwd` is canonicalized with filesystem semantics before becoming `workspaceRoot`, otherwise the configured fallback applies. Canonicalization precedes lexical normalization so `symlink/..` agrees with process working-directory resolution.
- `ctx.sandboxPolicy.defaultMode` / `ctx.sandboxPolicy.workspaceRoot` — the deployment default and fallback root used by `resolve()`.
- `effectiveSandboxMode(events)` — the pure fold of a session's `sandbox/mode` events (the last switch wins, or `undefined`), used inside `resolve()`.
- `setSandboxMode(session, mode)` — THE write path for a per-session override: appends exactly one `sandbox/mode` event. The switch IS its event; nothing mutates the mode out of band.

View File

@@ -17,12 +17,17 @@
import { resolve as resolvePath } from 'node:path'
import { Context, Service } from 'cordis'
import z from 'schemastery'
import type { SandboxExecutionPolicy, SandboxMode } from '@deepseek-ai/dsh-sandbox'
import { canonicalPath, type SandboxExecutionPolicy, type SandboxMode } from '@deepseek-ai/dsh-sandbox'
import type { Session } from '@deepseek-ai/dsh-session'
import { effectiveSandboxMode } from './session-mode.ts'
export { SANDBOX_MODES, effectiveSandboxMode, setSandboxMode } from './session-mode.ts'
/** Resolve filesystem identity before lexical normalization can erase symlink-sensitive components. */
function resolveWorkspaceRoot(path: string): string {
return resolvePath(canonicalPath(path))
}
declare module 'cordis' {
interface Context {
sandboxPolicy: SandboxPolicyService
@@ -80,7 +85,7 @@ export class SandboxPolicyService extends Service {
// runtime fact. `workspaceRoot` has NO schema default, so its fallback to
// the process cwd is real branching, resolved absolute either way.
this.defaultMode = config.mode as SandboxMode
this.workspaceRoot = resolvePath(config.workspaceRoot ?? process.cwd())
this.workspaceRoot = resolveWorkspaceRoot(config.workspaceRoot ?? process.cwd())
}
/**
@@ -96,7 +101,7 @@ export class SandboxPolicyService extends Service {
const { session } = request
return {
mode: request.mode ?? (session === undefined ? undefined : effectiveSandboxMode(session.events)) ?? this.defaultMode,
workspaceRoot: resolvePath(session?.header.cwd ?? this.workspaceRoot),
workspaceRoot: resolveWorkspaceRoot(session?.header.cwd ?? this.workspaceRoot),
}
}
}

View File

@@ -4,7 +4,9 @@
* override kit (fold + write path) both enforcing families read.
*/
import { resolve } from 'node:path'
import { mkdirSync, mkdtempSync, realpathSync, rmSync, symlinkSync } from 'node:fs'
import { tmpdir } from 'node:os'
import { join, resolve, sep } from 'node:path'
import { describe, expect, it } from 'vitest'
import { Context } from 'cordis'
import { Session, SessionId } from '@deepseek-ai/dsh-session'
@@ -67,6 +69,28 @@ describe('SandboxPolicyService', () => {
})
})
it('resolves a symlink-sensitive session cwd with filesystem semantics', async () => {
const root = mkdtempSync(join(tmpdir(), 'dsh-policy-cwd-'))
try {
const lexical = join(root, 'lexical')
const physical = join(root, 'physical')
const child = join(physical, 'child')
mkdirSync(lexical)
mkdirSync(child, { recursive: true })
const link = join(lexical, 'link')
symlinkSync(child, link, process.platform === 'win32' ? 'junction' : 'dir')
const cwd = `${link}${sep}..`
const ctx = await mounted({ mode: 'workspace-write', workspaceRoot: '/fallback' })
expect(ctx.sandboxPolicy.resolve({ session: session('sess-symlink-parent', cwd) })).toEqual({
mode: 'workspace-write',
workspaceRoot: realpathSync.native(physical),
})
} finally {
rmSync(root, { recursive: true, force: true })
}
})
it('lets an approved mode outrank the session mode while retaining its root', async () => {
const ctx = await mounted({ workspaceRoot: '/fallback' })
const active = session('sess-approved', '/projects/approved')

View File

@@ -6,7 +6,7 @@ The contract in one line: `ctx.sandbox.confine(argv, policy)` returns the argv t
Policy rides the call, not the provider: two consumers may confine under different policies at the same instant (bash under `read-only` while a confined child agent keeps its state directory writable), and an approved escalated retry is just a new call with a wider policy.
**Same-world confinement only.** A backend shares the host's filesystem and kernel (`bwrap`, Landlock, Seatbelt); `workspaceRoot` names a real host path. Containers, microVMs, and remote executors are NOT backends of this seam — they replace whole capability implementations (`ctx.bash`, `ctx.fs`) as environment-coherent groups. The boundary and its rationale: [the sandbox Agent Note](../../../.agents/notes/implemented/feature/2026-07-06-sandbox.md).
**Same-world confinement only.** A backend shares the host's filesystem and kernel (`bwrap`, Landlock, Seatbelt); `workspaceRoot` names the filesystem-canonical real host directory. Workspace identity is resolved before lexical normalization, so a valid cwd containing `symlink/..` grants the directory where `chdir` actually lands rather than an unrelated lexical parent. Containers, microVMs, and remote executors are NOT backends of this seam — they replace whole capability implementations (`ctx.bash`, `ctx.fs`) as environment-coherent groups. The boundary and its rationale: [the sandbox Agent Note](../../../.agents/notes/implemented/feature/2026-07-06-sandbox.md).
Implementations: [`@deepseek-ai/dsh-sandbox-local`](../sandbox-local/) (Linux: `bwrap`, else the per-platform Landlock launcher; macOS: `sandbox-exec`/Seatbelt). Consumers: [`@deepseek-ai/dsh-bash-sandbox`](../../bash/bash-sandbox/) (wraps `['bash', '-c', command]`).

View File

@@ -29,9 +29,13 @@ import type { SandboxExecutionPolicy } from './index.ts'
*/
export function canonicalPath(path: string): string {
try {
return realpathSync(path)
// Node's JavaScript realpath implementation lexically collapses `..`
// before resolving a preceding symlink on some platforms. The native
// implementation follows the filesystem's component-by-component lookup,
// matching chdir/spawn and the enforcement layers this identity feeds.
return realpathSync.native(path)
} catch {
// realpathSync failed: the path (or a prefix) is missing or unreadable.
// realpathSync.native failed: the path (or a prefix) is missing or unreadable.
return path
}
}