Local run of check:ci:windows-complete (the windows-native gate) on
latest master surfaced five Windows-only failures, all unreachable by
current CI because the native windows job is disabled and the wine gate
only covers build+site.
- install-lefthook/translation-pairing-merge specs junctioned the real
scripts/ and tsx package into fixtures; Windows recursive deletion
(Node rmSync and git worktree remove) follows MOUNT_POINT junctions and
deleted the repository's own directories mid-run. Fixtures now unlink
their reparse points before any recursive removal (shared helper in
scripts/test-fixture-cleanup.ts).
- workflow-workerthread spawned its worker with an empty env; on Windows
os.tmpdir() then degrades to the literal relative path undefined\temp,
so tsx wrote its transform cache into a cwd-relative undefined/
directory inside the repo. The worker env now injects the host temp
path on win32 (workerSpawnEnv, platform-parameterized and unit-tested
on both arms).
- workspace-context spec did not stub USERPROFILE (win32 homedir) or a
set DSH_HOME, leaking the developer machine's real ~/.dsh/AGENTS.md
into discovery.
- ui-trajectory client-bundle spec mounted the built artifact without the
remote/settingsScope provides the locale plugin needs, so the plugin
never activated and no view registered.
- subagent temp-fixture cleanup lacked the maxRetries Windows handle
release needs under load (EPERM); added retries to the three affected
specs and the fixture-cleanup helper.
The preset-roster loop added by this PR names the \cordis\ preset id in
apps/cli/tests/windows-shell.spec.ts; the bare token is product data like
the existing agent-preset allowlist entries, not a package reference, so
the rescope-vendor hygiene check must not rewrite it.
Drop the filler and factually loose comments added with the tool-pwsh rows:
the platform expressions self-explain, the web-app overlay comment repeats
the file header, and the spec comments now state only the platform-scoped
evaluation fact.
The web-app overlay now disables the host tool-pwsh row too, and the shipped
presets (standard/code/cordis) declare both shell tool rows with inverted
platform gates — tool-bash on POSIX, tool-pwsh on win32 — so the preset layer
exposes exactly one shell tool per host and a preset can drop or replace the
shell tool on either platform. windows-shell.spec pins both preset gates and
both host tool rows disabled in the web composition; the loader and Windows
pwsh notes are updated in place.
Address review: the gate module docstring now states the actual evaluation
contexts (config after injections against the plugin context, disabled at
every mount decision against the loader context); metadataExpressionErrors
rejects expressions nested below disabled and syntax-checks the disabled
expression itself so an unparseable gate fails at the gate instead of the
boot. The tutorial's !!js claims follow, and the loader note moves to
implemented/architecture with its inbound links retargeted.
Entry \disabled\ interpolation makes the launcher's separate platform layer
unnecessary: the base bundle's cordis.patch.yml now gates both shell stacks
on its own rows — bash-sandbox/tool-bash disable on win32, and their twins
pwsh-sandbox/tool-pwsh mount only there with the inverted expression — so
exactly one shell stack mounts per host from one shared patch file.
windows.cordis.patch.yml and the launcher's windows-shell.ts injection (boot,
live recomposition, config dumps) are deleted, with the workspace-constraints
entry and the dsh-base exports/files entries following. The windows-shell spec
pins the effective per-platform roster through the real bundle layers, and
base.spec pins the four symmetric gates. The superseded active notes are
updated and cross-linked; the loader note records the fold itself.
The Windows platform layer disables tool-bash and inserts the pwsh stack, but the shipped presets each mount a tool-bash row that re-enabled the tool on win32 — the session had both a PowerShell-backed bash tool and tool-pwsh, silently, because no spec pinned the composed preset layer.
The Loader now evaluates a disabled: !!js expression against the loader context at every mount decision; disabled is the only interpolated metadata field, and the raw node stays in the options so write-back keeps the !!js form. The standard/code/cordis presets gate tool-bash with process.platform === 'win32', verify-cordis-config allows expressions in disabled only, and the windows-shell spec pins the preset-level invariant.
The windows-native job has been red since #1990 put the sandbox-windows-acl sources into the Windows 100%-per-file denominator without tests carrying them, and #1543 dropped the authoring.ts V8 ignore for the POSIX-only owner-execute branch. Non-blocking at merge time, the red state has propagated to every later pull request.
Cover every in-process ACL-sandbox failure branch with stub-based failure-path suites (ffi/acl/token/spawn/index), following the package's existing failure-paths pattern; the package now measures 100% per file under the Windows denominator. Exclude only the runner entry from the win32 denominator: it executes exclusively as a spawned child outside the instrumented run, and its behavior is pinned end-to-end by the runner suite. Restore the authoring.ts narrow V8 ignore and add one for the dispose token guard whose absent-token arm is lifecycle-unreachable. Update the dual-lane Agent Note with the denominator composition.
Master's native Windows coverage lane now covers the LSP sources, and its
ci-workflow spec asserts the exclusion paths are absent from vitest.config.ts;
keeping them (as carried over in the merge) fails both the coverage and the
native Windows lanes. The sandbox-windows-acl exclusion stays: that package
is win32-only and the Linux lane cannot cover it.
The arrow shorthand implicitly returns rmSync's void, which the
no-confusing-void-expression rule forbids; the block body keeps the
fallback without the violation.
The durable sandbox/acl-session event carried a workspace binding that
always equals the session cwd and a random temp path that only needed
to be stable per session. Both are now derived: the temp subdirectory
is sha256(session id + workspace), created exclusively and removed on
provider dispose, so fork/resume semantics fall out of the derivation
and the record, its fold/provision/tamper validation, the immediate
flush kick, and the session-store dependency all disappear.
loadOverlayPatches already throws on a missing file (the caller named
it, so absence is a misconfiguration) — the existsSync guard was a
second fail-loud mechanism for the same miss with a prettier message.
The loader's throw keeps the fail-loud contract the Windows-default
note records.
The placer tests receive absolute paths; on Windows the POSIX-only
split kept the whole path as the basename. The windows-acl branch
already carries this fix from an earlier merge-forward.
The upgraded md-links gate resolves same-file fragments against the
zh file's own heading slugs; the translated heading needs the explicit
a-id anchor the corpus convention uses for cross-language fragments.
Under read-only and workspace-write the Windows ACL sandbox leaves
programs unable to open named pipes, so a piped-stdio spawn fails with
EPERM. State that boundary in the pwsh tool description next to the
ConstrainedLanguage contract, pin it in the tool tests, and bring the
package README and both implemented Agent Notes current with it.
CreateNamedPipeW with NULL security attributes does not install a kernel
template: KernelBase builds the documented 5-ACE default SD in user mode
and passes it down; the kernel itself (a raw SD-null create) applies the
token default DACL. Correct the claim in both README sides and the
runner.spec pin comment, and link the MS template documentation.
New objects created without an explicit security descriptor take
their DACL from the token's default DACL, which CreateRestrictedToken
builds from the user's ambient SIDs — none of them a restricting SID.
Confined children therefore failed the write pass-2 check when
creating anonymous pipes (CreatePipe: ERROR_ACCESS_DENIED, surfaced
as Node EPERM), breaking PowerShell pipelines and other CreatePipe
consumers. Merge a full-access write-SID ACE (Everyone under
read-only) into the token default DACL at init via
SetTokenInformation(TokenDefaultDacl).
Named pipes are EXEMPT: their default security descriptor is the
kernel's PUBLIC template (owner/SYSTEM/Admins full, Everyone
read-only), which no token change influences, so libuv's piped stdio
capture stays denied for confined grandchildren — the POC-documented
boundary, now pinned by the runner suite (inherit/ignore OK, pipe
DENIED) and documented in the README pair. The NUL paragraph is
corrected to the measured matrix (Everyone has 0x1201BF on the
device: cmd/node writes land; Set-Content fails at the PS layer).
The reparse case reuses the standing workspace grant of the preceding
case (same workspace -> map hit), so the failed temp grant is the third
grant, not the fourth; the dispose-warning text carries 'failure(s)'.
The per-session random write SID forced a full tree propagation per
session per server lifetime (minutes on large workspaces). The write
SID is now the per-workspace identity derived from the canonical
workspace path (workspaceWriteSid: sha256 -> S-1-4-x-y), stored
nowhere: the workspace-root ACE materializes once per workspace per
machine and every later provision hits the exact-ACE skip.
- workspace ACEs are STANDING (never revoked - the reuse cache); temp
ACEs stay revocable (disposed with the provider), so an inheritable
ACE never outlives its session's temp dir on the ambient temp root
- AclSandbox requires the write SID under workspace-write; read-only
parses/grants nothing; the runner derives the SID itself (the
--write-sid flag's presence still marks the seam-managed contract)
- the acl-session record drops writeSid (sessionId/workspace/tempDir
remain): the SID-tamper surface and its validation are gone
- sandbox-local holds two grant maps: standing workspace grants and
revocable per-session temp grants
Docs (README pair, design note pair, catalogs, type-equiv) and the
acl-session/grant/acl/probe/runner suites updated; workspace-sid.spec
pins the derivation contract.
The merge-forward took the parent's module-graph verbatim; the merged
tree adds the sandbox-windows-acl package and dsh-session edges.
verify-module-graph is a ci-static gate, not in local doc-sync.
The merge-forward collapsed the pwsh-sandbox key with the parent's
e2b entry into invalid JSON; restore the branch's original entry
alongside the kept e2b entry.
Master's pty-local rewrite dropped the sessionId expectation from the
resolved-policy contract; sandbox-policy resolve() injects the owner
sessionId (sandbox branch design). Restore it in both expectations.
Durable record: bound to the owning session id and validated at the fold (orphan-SID shape, temp path inside the host temp root) — a fork's copied parent record no longer provisions the child, and a tampered record fails loud. Private temp dir: random unguessable name persisted in the record, created exclusively (pre-existing entries and reparse points fail EEXIST). Persistence: a fresh provision kicks an immediate flush (no write-behind debounce), narrowing the crash window to the flush latency — documented as the one self-healing gap. Runner-failure rules: exit-gated on 127 so a confined command that prints the signature on a non-127 exit is never misclassified. Spawn: AssignProcessToJobObject failure terminates the suspended child (no hanging orphans). SandboxExecutionPolicy.sessionId is the branded SessionId. Boundary docs: qualifying clause on the absolutist sentences, NULL-DACL Known Limitation, 'full' scoped to the supported NTFS surface, CLM gate comment.
workspace-write now runs [logon SID, Everyone, orphan]: the two lists differ only by the orphan, and the keep-alive group (logon SID + Everyone) is the single shared invariant. The WMI namespace security check fails in BOTH modes (0x80041003), so CIM/Get-ComputerInfo are unavailable everywhere — the price of closing the C:\\-root tree-creation escape (AU:(AD) + AU:(OI)(CI)(IO)(M)) in workspace-write too. The unused WinLocalSid/WinInteractiveSid/WinAuthenticatedUserSid constants and their ABI-probe prints are removed; the enforcement 'full' claim now stands on a closed NTFS surface. New regression: a C:\\Users\\Public subdirectory write is denied under BOTH modes (the ambient-writable blind spot the review flagged — INTERACTIVE is absent from both lists). FAT-class (non-ACL) targets outside the granted roots remain writable (no security descriptors to intersect) — documented as a legacy residue, warn-only, not engineered around. Docs/design note/PR body updated in both languages (list I/J terminology gone everywhere).
tool-pwsh now renders sandbox denial facts (denial marker + same-turn escalation hint, runner-failed notice on background reads) and advertises the sandbox_permissions/justification escalation pair resolved through ctx.approval before execution — the shared fail-closed sequence from dsh-sandbox, with the tool name 'pwsh'. The description teaches the ConstrainedLanguage contract under the Windows sandbox (both confined modes run pwsh in CLM: Add-Type, non-core .NET statics, COM, and reflection fail; the mode cannot be lifted — probe-verified, pinned by LANGMODE assertions in the runner suite and description assertions in the tool suite). The stale pwsh-sandbox JSDoc ('the tool owns approval') and the parity notes' superseded 'minus sandbox' claims are corrected.
A session that materialized its workspace-write grant and then switched to read-only (or crashed and resumed under read-only) kept a writable workspace for the server lifetime: the standing orphan-SID ACE survived the downgrade and list I still carried the orphan, so the write-restricted pass-2 check granted it. List I is now [logon SID, EVERYONE] only — the standing ACE stays inert under read-only while the unrevoked ACE keeps the re-upgrade free (map hit, no re-propagation). Pinned by the runner's real-token mode-downgrade regression (standing ACE denied under read-only, effective again on re-upgrade), the acl-session mode-switch cycle (nothing under read-only, one materialization on upgrade, no revoke on downgrade), and ConstrainedLanguage pins in both modes.
quoteArg doubles end-of-string backslashes (real CommandLineToArgvW round-trip test; first-token exemption noted); grantWrite merges into the current DACL instead of replacing it, and both grant/revoke hold a LockFileEx per-path lock under GetTempPathW/dsh-acl-locks (koffi crashes on NULL lpOverlapped — a zeroed OVERLAPPED is used); GRANT_MASK gains DELETE|FILE_DELETE_CHILD (never WRITE_DAC/WRITE_OWNER) though the win32 26200 second check only constrains the WRITE bit; failure paths close all handles (CreateProcessAsUserW pipe set, ResumeThread thread/process/job) with stub-api tests; getTempPathW refuses undersized buffers; drainPipe backs off; runner.spec pwshAvailable uses the resolvePwshPath probe; NTSTATUS exit codes mirror bit-exact (verified end-to-end); WinLocalSid JSDoc re-attributed to WinLocalLogonSid. All gates green: 42 passed/2 skipped, typecheck, oxlint, 0 clones, constraints, knip.
Known Limitations gains the per-command DACL mutation entry (inheritance is lazy, not a per-file walk; per-session grant reuse is deferred work); the design note gains a Testing section stating why the win32-only roster flip cannot ride macOS/Linux snapshot fixtures and naming the substitute evidence.
The launcher's healProfilesModuleFallback BFS-links the apps/cli dependency closure into the profile's node_modules (the pwsh-local precedent): without dsh-pwsh-sandbox in apps/cli dependencies, a fresh Windows host cannot resolve the inserted row. Assert the closure reaches both inserted packages in the real-layers composition; rewrap the windows-shell layer comment.
The duplication gate flags the call-for-call mirror of bash-sandbox's executor (the pwsh-tool-and-executor decision): wrap the class body in jscpd ignore markers, the same convention the invariant companions use.
sandbox-local now depends on @deepseek-ai/dsh-sandbox-windows-acl, so the packed-install e2e must carry its tarball in the workspace closure — npm cannot resolve the private package from the registry. koffi and the shared peers (dsh-invariants, cordis) already resolve from the registry or the existing closure.
The self-hosted Linux runners ship no pwsh, and the pwshAvailable probes used spawnSync('where.exe'), which reports a missing binary as status null instead of throwing — the suites never skipped and failed with spawn pwsh ENOENT. Probe with resolvePwshPath() status instead, the same gate the coverage exemption uses. Exempt pwsh-sandbox src from coverage on pwsh-less hosts (its remaining helpers branch and the invariant companion ride the executor suites' real pwsh runs); pwsh-ful hosts keep the full 100% bar. Cover the windows-acl probe case and the runner-entry resolution in sandbox-local on Linux (chain-seam tests plus a windowsAclRunnerEntry seam) — the package's POSIX-only suites are Linux's only chance to cover the new lines. Static gate fixes: declare dsh-pwsh-sandbox in the base bundle, register the runner files entry in constraints, knip entries for the e2e suite and where.exe, regenerate the module graph. Verified in WSL (no-pwsh Linux): pwsh-sandbox 5 pass/13 skip with the exemption active, sandbox-local coverage 100%.
Add the mechanism section to the Windows rung Agent Note: identity routes (restricted-user, AppContainer) default-deny everything from a fresh SID and pay per-path ACE writes for every read, while the WRITE_RESTRICTED token keeps the caller's SID and double-checks write-class access only — reads pass on the normal check alone. Covers DISABLE_MAX_PRIVILEGE/LUA_TOKEN as the token-side limited-user synthesis and why SidsToDisable read restriction is unused.
Add the implemented Agent Note owning the win32 rung decision: mxc rejected (OS floor 24H2, BaseContainer tier only on 25H2+ with the OS feature enabled, and arbitrary-path reads demand wholesale host DACL writes), AppContainer rejected (no ambient read access — arbitrary-path reads unsupported), landstrip already rejected and AppContainer-shaped. Link the note from the backend README pair and restate the choice in the PR description.
JSDoc every sandbox-windows-acl export (the 138 verify-export-jsdoc violations); fix the sandbox-windows-acl README code block (undeclared workspaceRoot), bring its stale claims current (the provider wiring and kill-on-close are live), add the canonical Model Experience sections to both new packages, audited the backend's indirect Model Experience sentence, and regenerate docs/config-catalog.md.
The placer fixtures derive the basename with split('/'), which never splits a Windows absolute path (backslashes), so the two image-placement tests fail on win32 hosts. The production rewrite passes absolute paths through; the fixtures now accept both separators.
The Windows platform layer re-enables the base fs-sandbox row (removing its disable), but still inserted dsh-fs-local: both extend FileSystem and provide ctx.fs, so every shipped win32 profile failed at load. Delete the insert; fs-sandbox stays the single fs provider exactly as on POSIX. Sync the roster specs, the base/reference README pairs, the sandbox core doc (read-only grants no sink on Windows), the windows-shell JSDoc, and re-record the i18n pairings.
Confine Windows command execution through a WRITE_RESTRICTED token whose restricting SIDs carry an orphan-SID write allowlist, ported from https://github.com/huoyaoyuan/windows-acl-restrict-poc (@ 10e4dfb). Every Win32 call is checked and fails closed - the POC silently ran children with the FULL token when CreateRestrictedToken failed.
- @deepseek-ai/dsh-sandbox-windows-acl: koffi primitives verified against the MinGW Windows headers (verify/abi-probe.cpp) plus the confinement runner ([node, runner, --workspace, --temp, --mode, --, argv...]: kill-on-close job, stdio passthrough, exit-code mirroring, windows-acl-run: failure signature, grant revocation). read-only = strict zero grants (NUL device not writable; documented). Windows-only execution: exempted from the Linux coverage lane (windowsOnlyCoverageExclusions).
- @deepseek-ai/dsh-sandbox-local: PLATFORM_CHAINS.win32 filled with the windows-acl runner (full enforcement, ACL denial dialect, runner-failure rules).
- @deepseek-ai/dsh-pwsh-sandbox: sandbox-consuming pwsh executor (call-for-call mirror of dsh-bash-sandbox) over a new argv-level seam in dsh-pwsh-local; per-file coverage complete via the fake-provider spec.
- bundle/base: the Windows platform layer mounts the confined pwsh roster - sandbox/policy/fs-sandbox/permission/approval re-enabled, the POSIX bash stack stays disabled.
Co-authored-by: Huo Yaoyuan <huoyaoyuan@hotmail.com>
Huanqi Cao
