7 Commits

Include renames

Author	SHA1	Message	Date
garrytan	f054927ee3	Format lib/cso with prettier	2026-10-08 05:53:29 +00:00
garrytan	5a0dcb2bab	fix(cso): Python and Rails repairs verify and replay; killed attempts leave no scratch ... - Python offline preparation sets SOURCE_DATE_EPOCH so .pyc bytecode is reproducible. - Prepared identity hashes skip RubyGems/mini_portile build diagnostics that name random build dirs. - Rails verification attempts and replays get 900 s; the attempt watchdog, Docker verification executor and assertion witness share MAX_VERIFICATION_ATTEMPT_MS. - resume and finish reclaim a killed attempt's scratch once its watchdog recorded cleanup. v1.91.34.0	2026-10-07 20:33:58 +00:00
+818 people	28f1385eac	v1.91.32.0 fix: Codex works from zsh, free text never enters shell commands, honest failures (#3057) ... * fix: sourced helpers locate themselves under zsh (#3024) Skill blocks source bin/gstack-codex-probe and bin/gstack-egress-lib.sh into the host's shell. Under zsh (macOS default) BASH_SOURCE is empty, so Codex model selection resolved /../scripts/... and every outside pass reported CODEX_MODEL: invalid; the egress lib resolved to the caller's cwd. Resolve the file once at source time (BASH_SOURCE under bash, %x under zsh), and replace the bash-only ${!name} indirection in the auth probe. A free test derives every sourced helper from the generated skills and sources each under bash and zsh. Co-authored-by: marcosdetry <marcosdetry@users.noreply.github.com> Co-authored-by: akshaykalia <akshaykalia@users.noreply.github.com> Co-authored-by: scottgigante-hubflow <scottgigante-hubflow@users.noreply.github.com> Co-authored-by: woodsonl <woodsonl@users.noreply.github.com> Co-authored-by: bezdekt <bezdekt@users.noreply.github.com> * fix: reply and issue templates never pass untrusted text in double quotes (#3046) greptile-triage's reply APIs posted -f body="<reply text>" while the templates quote SHAs and diff code in backticks; in double quotes bash ran each span (replies posted as 'Fixed in .', and diff or reviewer text ran as a command). The pre-existing-failure issue template and /spec's issue title had the same shape. All three now go through quoted heredocs. The generated-bash lint gains untrusted-in-quotes and now also covers runtime reference docs in skill directories (review/greptile-triage.md). Co-authored-by: dgrant <dgrant@users.noreply.github.com> * fix: a Codex usage limit is its own state, not 'inconclusive; proceeding' (#3051) The model probe matched only model and broken-install errors, so a usage or rate limit fell through to MODEL_PROBE_INCONCLUSIVE: Codex's own message (with the reset time) was dropped, the hint pointed at model config, and every later skill re-probed for 30 s. The probe now returns MODEL_QUOTA_EXHAUSTED (exit 4) on a failed, non-timeout call that names a usage limit, rate limit, insufficient_quota or 429, relays Codex's line verbatim and caches it for 15 minutes. Preflights map it to CODEX_MODE: quota_exhausted (missing coverage, no further Codex calls), and classifyOutsideReview reports a mid-run limit as unavailable (quota) with its own troubleshooting anchor. Co-authored-by: venkatareddyg <venkatareddyg@users.noreply.github.com> * test: record ship/review union sizes after the quota branch and issue template * fix: Codex watchdog freezes the command before reaping its children The bash-native watchdog (stock macOS, no timeout binary) ran pkill -P on the command's children and then killed the command. Between the two, the command continued past its killed child and printed output after its deadline. Stop it first. The test's minimal PATH also lacked pkill, which the watchdog uses to reap children (stock macOS ships it), so the orphaned child held stdout open and the capture waited out its sleep. * fix(autoplan): an oversize session journal is a hard too_large denial, not a retryable identity failure (#3050) The guard capped journal reads at 32 MiB and reported a longer journal as 'identity', which the hook treats as deny-and-retry. A journal only grows, so /autoplan could never enter Phase 1 in a long session, and the troubleshooting row sent people chasing symlinks. Oversize journals now report too_large as a hard cause that names the size and the supported fallback (run the reviews by hand or start a new session). Co-authored-by: venkatareddyg <venkatareddyg@users.noreply.github.com> * fix(deps): clear the weekly OSV findings (#3040) Raise overrides to fixed releases: ip-address 10.7.3 (our own pin was the vulnerable 10.3.1; reachable through socks in browse's proxy path), fast-uri 3.1.8, hono 4.13.13, proxy-addr 2.0.8, @modelcontextprotocol/sdk 1.31.0, sharp 0.35.5, and in lib/diagram-render dompurify 3.4.16 (bundled by mermaid) and source-map-js 1.2.2; the render page is rebuilt. Three advisories without a usable fix get justified, expiring ignores: braces (build-time only, excalidraw -> sass -> chokidar; a test checks the page carries no chokidar code), sprintf-js (roarr's constant-format logger) and katex (needs pre-existing prototype pollution; fix outside mermaid's range). osv-scanner with the repo config: no issues found. * fix(autoplan): name a shell-built snapshot init as the cause of the missing-evidence denial (#3045) The guard binds a phase entry only to a literal snapshot-init Bash call. When the model resolved the tool path into a variable, chained or piped the call, init ran and succeeded but the denial said 'Complete the existing snapshot init step', which reads as already done, so the run gave up. When no init binds but a Bash call looks like one, the denial now says the shape cannot be bound and prints the literal re-run (answers reused:true). The binding rule is unchanged, and the skill says to run init on its own. Co-authored-by: CharlesBerg85 <CharlesBerg85@users.noreply.github.com> * fix(browse): chain reads piped stdin synchronously, so Windows runs the flow (#3039) On Windows, an awaited Bun.stdin.text() inside the CLI's un-awaited main() let the process exit 0 before reading stdin: a replayed flow reported success with nothing sent to the daemon. Read fd 0 synchronously (skipped for a terminal); empty input now gets the server's usage error and a non-zero exit. Co-authored-by: VaibDeveloper <VaibDeveloper@users.noreply.github.com> * fix(browse): clearer config-mismatch refusal and a WebGL note for the macOS GPU flags (#3030) A plain call refused because another session started the daemon with --headed/--proxy was told to 'apply --proxy/--headed', flags it never passed. It now says who started the daemon and offers both remedies. The BROWSER.md env table notes that the default macOS GPU flags disable WebGL and that GSTACK_DISABLE_GPU=off restores it for WebGL QA. Co-authored-by: Hagetino <Hagetino@users.noreply.github.com> * fix(ship): skip Greptile triage when no PR exists yet (#3020) /ship's first run on a branch dispatched a triage subagent only to learn there was no PR. Step 10 now checks for a PR inline (gh's 'no pull requests found') and records that Greptile runs on the PR once it exists; other gh errors still dispatch so the child can report unavailable. Co-authored-by: dgrant <dgrant@users.noreply.github.com> * fix(router): route only to loaded skills, and say when not to invoke (#3018) The root router listed every skill and ended 'When in doubt, invoke the skill'. Current Claude models follow that literally: they tried to invoke skills a user had turned off in Claude Code's skillOverrides (which gstack's own disabled_skills render cannot see), and borderline requests tipped into full workflows. The router now routes only to skills in the session's available-skills list and names when to answer directly; the CLAUDE.md routing block skill-start offers says the same. Co-authored-by: Rekin226 <Rekin226@users.noreply.github.com> * fix: spawned sessions never auto-grant consent or publish off the machine (#3016) Spawned sessions auto-choose the recommended option except for destructive or irreversible choices. Main already withholds the artifacts-sync consent prompt from spawned sessions; the remaining gap was any other prompt whose recommended option grants a consent or syncs data off the machine. Every spawned directive (skill-start's block, the escape sentence, the Conductor deny reason and the AUQ-error fallback) now names that as irreversible. A consent the user already gave interactively stays in force. Co-authored-by: ekson73 <ekson73@users.noreply.github.com> * test: fix the spawned-block slice and type the #3045 re-init events * fix(redact): TypeScript types and JSX expressions are code, not env.kv secrets (#3048) A credential-named parameter or prop with an unquoted code value (`session: SessionState,`, `key={turn.requestId + turn.role}`, `apiKey={settings.apiKey}`) cleared the entropy gate and blocked pushes as a MEDIUM env.kv finding. One rule replaces the PR's two keyword heuristics: a value that reads as a code name (no digits, word-like camelCase or dotted parts) with a code terminator, or a brace expression of such names with no quoted literal, is source syntax. Paired controls keep firing on quoted, brace-wrapped, concatenated, YAML-bare and random-letter literals. Co-authored-by: tianxchen <tianxchen@users.noreply.github.com> * fix(setup): refuse to register a hook that does not parse (#3032) Claude Code runs gstack's hook shims through /bin/sh; a shim that does not parse exits 2, which for a PreToolUse hook blocks the tool call in every session until someone finds the file. Before any hook registration is healed or written, setup now runs bash -n on each registered shim and a Bun parse of its TypeScript entry, and stops with the file, line and fix. A lean rewrite of the PR's idea (one gate block plus a test that runs it). Co-authored-by: BenjaminDSmithy <BenjaminDSmithy@users.noreply.github.com> * fix(browse): Windows runtime roots start the server from the source checkout's bundle (#3026) On Windows a host runtime root (~/.codex/skills/gstack and the other env-var hosts) holds file copies of browse/dist and no node_modules, so its server-node.mjs could not import playwright and every browser skill died (the module-scope server.ts half was fixed in v1.91.19.0). setup now records the source checkout in every runtime root's .source-path, and the CLI runs that checkout's bundle when the adjacent one cannot reach playwright. Verified by unit layout tests and the Windows CI lane, not live on Windows. Co-authored-by: ubay887 <ubay887@users.noreply.github.com> * fix(copilot): on Windows, run bash blocks in Git for Windows Bash (#3047) Copilot CLI on Windows ran gstack's extensionless helpers straight from PowerShell, which opened an app picker, and bare bash can resolve to WSL. The Copilot tool glossary (rendered into every Copilot skill) now says to run each bash block in Git for Windows Bash by its full path and never to translate a block into PowerShell. The PR's PowerShell repair skill and README section are not taken. Co-authored-by: nareshh74 <nareshh74@users.noreply.github.com> * test: record ship's union size after the Greptile no-PR check * fix: keep issue numbers out of model-read Copilot text; seed the new routing sentence in the autoplan fixture * docs: Oct 6 fix-wave plan, reviewed through /autoplan * fix(autoplan): every oversize-journal path reports too_large with a recovery that keeps the work Review follow-up to the #3050 fix. The summed batch read and the narrowed parent loop now report too_large with the real size instead of a plain error naming a hard-coded 32 MiB. The denial names the size and the cap, and gives the recovery that keeps work (/context-save, new session, /context-restore, /autoplan) with a docs anchor. A test seam that can only lower the cap replaces the 33 MiB fixture. * test(autoplan): negative controls and more shapes for the shell-built init denial * fix(redact): decide TS/JSX source syntax by file context, not by value shape Review follow-up to the #3048 fix. The word-shape heuristic also exempted real secrets made of words (API_KEY=VelvetRiverOrbitSunset; in a .env). The scanner now takes the file path (ScanOptions.sourcePath): the pre-push hook scans the pushed diff per file with its +++ b/ path, and gstack-redact --from-file passes its path. Only in a TypeScript/JSX file is an unquoted value after ':' or a JSX brace of names treated as source; the same line in .env/YAML, or with no path, is scanned as a literal, and the jwt pattern still fires in .ts files. Co-authored-by: tianxchen <tianxchen@users.noreply.github.com> * docs: gstack's own CLAUDE.md routing block matches the new router wording Live check (Claude Code 2.1.284): a skill set to "off" in skillOverrides is absent from the session's skill list while its sibling is listed, and both are listed without the override, so 'route only to skills in your available-skills list' holds. * fix(sync-gbrain): don't offer a call-graph build for a source with no code pages (#3029) Step 3.5 offered --dream for any source with pages, including one that holds only markdown (synced with gbrain's default strategy), where a build cannot help, and the injected CLAUDE.md guidance told future sessions to run --dream for any count: 0. On gbrain >= 0.60, code-def for an impossible symbol answers .status (probed live against gbrain 0.60.96: ready for a code source, out_of_scope for a markdown-only one). Step 3.5 now probes the pinned source first, reports the cause instead of asking on out_of_scope, and keeps today's offer when the probe is unknown (older gbrain, an error). gstack's own sync uses --strategy code, so the PR's soft-delete claim does not apply to gstack's path. Co-authored-by: syong <syong@users.noreply.github.com> * fix(setup): register the hooks that parse, refuse the rest with file:line (#3032) The landed gate stopped setup before any registration, so one unparseable hook left every other hook unregistered. On the Claude install the canonical root is the checkout Claude Code runs hooks from, so there is no staged swap to abort: setup now registers the hooks that parse, refuses the ones that do not with file and line, finishes the install and exits non-zero, saying it is a gstack bug and that a skipped hook keeps running from the checkout. bin/gstack-hook-check derives its list from setup's _hook_command_path calls plus the autoplan phase-publication hook (a test keeps them equal), parses each shim with its shebang's interpreter, and bundles each TypeScript entry with its local imports (packages external) into a temp dir, so a broken or missing import, or a missing entry, fails. Co-authored-by: BenjaminDSmithy <BenjaminDSmithy@users.noreply.github.com> * fix(auto-update): hold an update whose hooks do not parse, before the checkout moves bin/gstack-session-update fast-forwards the live checkout before setup runs, and Claude Code runs hooks straight from it, so setup's gate cannot protect an auto-update. After the Bun floor check the updater extracts the incoming revision with git archive and runs bin/gstack-hook-check on it; a failure holds the update like bun-too-old, records hook-does-not-parse with file:line in the update status, and the next session start says the current hooks keep running. Fixture-remote tests cover a broken shim, a broken imported module and a parseable control. Co-authored-by: BenjaminDSmithy <BenjaminDSmithy@users.noreply.github.com> * fix(browse): chain with no flow is a usage error before any daemon boots (#3039) Empty piped input, a terminal with no arguments, and a readFileSync(0) error (EAGAIN, EOF) now print the chain usage with the cause and its troubleshooting link and exit 1 before ensureServer(), instead of sending an empty argument to a daemon the CLI booted for it. Tests run the real CLI with no daemon and no browser, plus the stub-daemon flow. Co-authored-by: VaibDeveloper <VaibDeveloper@users.noreply.github.com> * fix(browse): refuse a runtime-root CLI and checkout server from different builds (#3026) A Windows runtime root runs the source checkout's Node server bundle through .source-path. When the root's CLI and the checkout's bundle carry different browse/dist/.version builds (the checkout was rebuilt without refreshing the root), the CLI now refuses to start the server with both builds, the ./setup fix and its troubleshooting anchor. Unversioned dev builds are not called skew. Co-authored-by: ubay887 <ubay887@users.noreply.github.com> * ci(windows): real isolated ./setup install, then a piped chain flow through the installed browse windows-setup-e2e built the checkout but never ran ./setup. It now installs the codex host into an isolated HOME (a runtime root copy with no node_modules), then from an unrelated directory pipes a two-command chain flow at a loopback page into the installed browse and asserts the page text, and that empty piped input exits 1 with the usage error. Covers the checkout-bundle path and the piped-stdin path end to end. * test(lint): free text never enters a generated command (structural rule) The generated-bash lint's untrusted-in-quotes rule (gh/glab/git/curl text flags in double quotes only) becomes free-text-placeholder: any <placeholder> anywhere in any command, in double, single or no quotes, and in heredoc bodies (quoted or not), unless it is on the reviewed identifier allowlist (test/helpers/placeholder-allowlist.ts) with the grammar a value must match. Quoted-only identifiers (URLs, selectors, paths printed by gstack) must sit inside quotes. Backslash-newline continuations now read as one command, so a flag and its value on separate lines are still seen together. Co-authored-by: dgrant <dgrant@users.noreply.github.com> * fix: reply, issue and spec text travel as agent-written files, never through a heredoc Replaces the quoted-heredoc bodies with CEO-12's file mechanism: a bash block creates a private mktemp file in the project's .gstack/tmp/ (kept out of git through .git/info/exclude), the agent writes the text with its file-write tool, and the next block posts with -F body=@file, --body-file or "$(cat file)". A heredoc ends early at any body line equal to its delimiter; a file has no such line. When the file was never written the block refuses with the path and the command to send it by hand, and it deletes the file after a successful post. - review/greptile-triage.md: replies; the fetched JSON moves from fixed /tmp/greptile_*.json paths to a mktemp directory; REPO and PR are re-derived per block instead of read from a previous shell. - Test-failure issues: title and body files, platform case on github/gitlab, assignee only when it is a valid login. - /spec: one draft file feeds the audit hash, every redaction scan, the issue and the archive; the archive no longer expands <title>/<body> inside an unquoted heredoc; the decision-log JSON is built by jq from files. Location verified against Claude Code 2.1.284 (Dockerfile.ci pin): project-local .gstack/tmp is writable under acceptEdits with no prompt and keeps mode 0600; $TMPDIR and the gstack state root prompt even under acceptEdits. Co-authored-by: dgrant <dgrant@users.noreply.github.com> * fix(ship): Greptile triage runs only when a PR exists; every other case says why PR: unknown (gh missing, not logged in, a non-GitHub remote, or a failed lookup) used to dispatch the triage child anyway. The Step 10 check now tests gh, its login, the remote and the PR in that order, and every case but 'PR: exists' skips the dispatch and records 'Greptile: not run (<reason>); runs on the PR once it exists'. Co-authored-by: dgrant <dgrant@users.noreply.github.com> * fix: question-tuning commands carry no free text The question log took the summary in single quotes and the preference write took the user's own words; an apostrophe ended the quote and the rest ran as shell. The summary is now a kebab-case slug (an identifier with a grammar), and the optional free_text field is no longer passed from skills (the binary still accepts it). * fix(cso): Windows finds docker.exe only under known-folder roots and says transport is unsupported (#3028) On Windows /cso looked for docker only in Git's directory and System32, so Docker Desktop was never found and the gap read as 'not installed'. It now resolves Program Files, Program Files (x86) and the Windows directory through PowerShell's [Environment]::GetFolderPath (not the process environment), checks each candidate's real path against those roots and rejects reparse points. lib/cso/docker.ts admits only unix:/// endpoints, so a trusted docker.exe reports 'native Windows Docker transport is not supported yet; static assessment only'; one outside the roots is refused as outside the trusted install locations, not overridable. Both carry an anchor. Native named-pipe transport is not implemented. * fix: sourced helpers share one self-locate block and fail loudly when they cannot locate themselves (#3024 rework) The codex probe and the egress lib now carry one byte-identical block: BASH_SOURCE under bash or %x under zsh, made absolute with $PWD, else a set GSTACK_ROOT that holds the helper. Anything else (dash, sh, a shell that cannot name the file) prints "gstack: cannot locate <helper> (shell: ...)" with the docs/troubleshooting.md#sourced-helper-location URL and returns before any bash- or zsh-only syntax. The Claude-only default path and the $(pwd) fallback are gone. The generated Codex preflight dot-sources the probe and, when it cannot load, reports CODEX_MODE: helper_unavailable with the helper's own line instead of falling through to not_authed. The class guard enumerates nested sources, makes real resolver, auth, state-root and egress-home calls under bash and zsh (absolute, relative then cd, bare name via PATH), covers dash and an unknown shell, runs the full preflight, and lints BASH_SOURCE to the shared block plus other bash-only constructs. Co-authored-by: marcosdetry <marcosdetry@users.noreply.github.com> Co-authored-by: akshaykalia <akshaykalia@users.noreply.github.com> Co-authored-by: scottgigante-hubflow <scottgigante-hubflow@users.noreply.github.com> Co-authored-by: woodsonl <woodsonl@users.noreply.github.com> Co-authored-by: bezdekt <bezdekt@users.noreply.github.com> * fix(cso): a Maven or Gradle repository is Java, not Python (#3028) With no supported stack metadata, preparation defaulted to Python and asked a Java repository for requirements.txt. A root pom.xml, build.gradle or build.gradle.kts now names the stack as Java (unsupported for runtime preparation, so the prerequisite says so), and Python helper scripts or a root requirements.txt beside it no longer make it Python. * fix(browse): BROWSER.md says plainly how to enable WebGL; the mismatch refusal links Daemon discipline (#3030) The GSTACK_DISABLE_GPU row now states the effect and the switch: set GSTACK_DISABLE_GPU=off to enable WebGL (off is the only recognized value). The reworded config-mismatch refusal links BROWSER.md's Daemon discipline section. Free tests pin both. Co-authored-by: Hagetino <Hagetino@users.noreply.github.com> * test(windows): name the B2, B4, B5 and B12 tests in the curated Windows lane The runtime-root bundle tests lived in browse/test/config.test.ts and the Copilot glossary test in test/host-config.test.ts, both excluded from the Windows lane, so the two Windows-only fixes were never checked on Windows. Move each into its own Windows-safe file (unchanged assertions) and list it, the piped-stdin chain test and the /cso docker.exe discovery test as named Windows coverage in scripts/lib/windows-curation.ts. * fix: the free-text block exits loudly when mktemp fails; spec and redact tests follow the file mechanism The redact resolver test, the #2679 mktemp guard test, the /spec invariant and the spec secret-sink harness pinned the heredoc transport. They now pin the agent-written file: the harness runs the draft-file block, writes the draft the way the file-write tool would, then runs the real scan fence, with every HIGH/MEDIUM/scanner-failure assertion unchanged. Co-authored-by: dgrant <dgrant@users.noreply.github.com> * test: hostile payloads through the generated posting blocks, identifier grammars, every quoting shape - The lint reads a whole command, so an unquoted multi-word placeholder is caught; negative controls cover double, single and no quotes, quoted and unquoted heredocs, continuation lines, and quoted-only identifiers used unquoted. - Every allowlisted grammar rejects quote-breaking characters (and shell operators and whitespace when used unquoted); git accepts 'fix;id' and 'fix`id`' as branch names, the ref grammar does not. - Greptile replies, test-failure issues (GitHub and GitLab) and /spec filing and archive run from the real generated text with stub gh/glab: a body with backticks, $(...), an apostrophe and lines equal to EOF and the old heredoc delimiters arrives byte for byte, nothing in it runs, and the temp file is gone. An unwritten file is refused with its path and the manual command, and nothing is sent. Co-authored-by: dgrant <dgrant@users.noreply.github.com> * test(ship): the PR check precedes the Greptile dispatch, and each skip names its reason Runs Step 10's generated check against stub gh for gh missing, not logged in, a non-GitHub remote, no PR, a failed lookup and an existing PR. Fails on the previous text (which printed PR: unknown and dispatched). Co-authored-by: dgrant <dgrant@users.noreply.github.com> * feat(ship): ship-measure runner for the measure-then-fix loop (A6) scripts/ship-measure.ts owns /ship's diagnostic measurement of a red eval case: per-kind aggregation (rule trials; behavior panels of 3, each at its own 2 of 3 plus 90% of trials; judge outputs scored by 3-sample panels, standalone judge ids accepted), one artifact directory per baseline, repair round and trial, a diagnostic label with verdict null that refuses the eval history dir, an estimated admission budget that reserves every concurrent trial and reconciles actual cost, the ask-once rule, the repair-round limit, the unmeasured skip and the per-case PR table. Free-suite flakes (CEO-1): 'free' reruns one shard's exact file list N times, rerun 1 alone as the baseline and the rest in parallel at max(1, floor(cores / shard concurrency)), each with its own HOME, state root and flake ledger, GSTACK_FREE_RETRY_FLAKY=0, under a wall cap that reports a partial count. test:ubicloud --diagnostic runs the same on one VM with the retry off. bin/gstack-config gains the DX-7 keys with their CEO-9 defaults and rejecting validators. Free tests drive every kind, the budget stop, the unmeasured skip and the free reruns with fake runners. * fix: a Codex usage limit is quota_exhausted, a plain 429 is rate_limited, and only Codex's own error lines count (#3051 rework) The probe's quota signature used to treat any rate limit, "too many requests" or 429 as quota, and read the whole output, which echoes the prompt. Now only Codex's trailing ERROR: / stream error lines are read. A usage-limit or insufficient_quota line is MODEL_QUOTA_EXHAUSTED (exit 4, cached 15 minutes; quota wins over a 429 that names insufficient_quota); a plain 429 is MODEL_PROBE_RATE_LIMITED (exit 0, state rate_limited, never cached), reported as CODEX_MODE: unverified (rate_limited) so the review still runs and decides. The quota HINT says how many minutes gstack skips Codex and how to retry now (GSTACK_CODEX_PROBE_RETRY=1 skips the negative cache for that probe, or delete <state root>/.codex-model-probe), with the full troubleshooting URL. classifyOutsideReview reads the same trailing lines: a mid-review quota is unavailable (quota_exhausted), a mid-review 429 unavailable (rate_limited). New gate row rate_limited and docs anchor codex-rate-limited; the quota row documents the retry. Tests cover each line shape, the echoed prompt, exit 0, timeout, retry recovery, probe/classifier parity, and two fresh-shell preflight blocks where the second makes no paid Codex call. Co-authored-by: venkatareddyg <venkatareddyg@users.noreply.github.com> * ci: the eval image installs zsh, and the macOS lane sources the Codex probe from zsh Dockerfile.ci adds zsh so the sourced-helper guard runs (and fails loudly when zsh is missing) inside the eval container too. The macOS named regressions job sources bin/gstack-codex-probe from zsh in an unrelated directory and runs _gstack_codex_select_model exec (no paid call), then runs the sourced-helper guard on the macOS runner. A free test pins both lanes and runs the macOS step's script locally. * test: deterministic regression for the Codex watchdog freezing the command before reaping its children The existing stubborn-child test caught the reap-then-kill race only some of the time (2 of 3 runs on the pre-fix watchdog). A pkill that sleeps after reaping widens the window between reaping the children and killing the command, so a watchdog that does not SIGSTOP the command first always lets it print "late" after the deadline. Fails 3 of 3 on the pre-fix watchdog, passes 3 of 3 on the current one. * test: live codex smokes run codex in a private HOME and CODEX_HOME codex-resume-flag-semantics and codex-e2e-sol-scope probe the codex CLI (the latter at module load, in every free run), and every codex run writes $CODEX_HOME/tmp/arg0, so with a real codex on PATH they wrote into the developer's ~/.codex. Both now give codex a private HOME and CODEX_HOME and remove it afterwards. A free test runs each file against a stub codex that records HOME and CODEX_HOME and asserts neither is the caller's. * feat(ship): a red eval runs classify, measure, fix, re-measure, gate once (A6) Step 6 no longer stops at the first red paid eval: it reads the new lazily loaded measure section and runs the loop for each red case, and Step 5 measures a failed free-suite shard (diagnostic reruns, retry off) before the triage can call it pre-existing. The section states the diagnostic- not-verdict rule, the per-kind table, the ask-once and named-red stops, the unmeasured label and the honest-fix wording (observed k/n after fix at <cause>). A free test pins the steps and their order in the generated skill and checks every command and flag it names exists in the runner. Ship's parity cap rises to 1.455 (measured 272,944 / 187,706 = 1.4541). * docs(testing): diagnostic measurement is not a verdict; per-kind measure table (A6) * test(ship): seeded-flake paid case for the measure loop (A6, periodic, behavior) A fixture repo with no remote documents a single-case eval command and a gate command, both free local stubs. queue-priorities claims trial slots with atomic mkdir (reset per ship-measure round) and fails slots 2, 5 and 9 while src/priorities.js leaves its seed-shuffled list unsorted, printing failure_cause and failure_detail that name the line. The paid case gives /ship's generated Step 6 and measure section to a live session and asserts it measured the case alone, fixed the named line without touching the eval, re-measured at least 10 passing trials on the sorted file, and only then ran the gate once; any question other than the spend question is a contract violation. A free test measures the fixture through the real ship-measure CLI: 7/10 unsorted, 7/10 after an unrelated edit with the gate still red (ENG-13's negative control), 10/10 after the sort with the gate green. * test(cso): the missing-scanner case points Docker at a dead socket 'an optional scanner prerequisite does not turn completed manual coverage partial' assumed no usable Docker. On a machine with Docker and a pulled gitleaks image the scan completed and the test failed (also on main). The scan step now gets a DOCKER_HOST that cannot connect, which is the prerequisite-missing condition the test describes. * fix: Aside prompts travel as agent-written files, never inside the command The open-ended reading recipe (browse, devex-review), scrape's fuzzy-intent read and the shared web-research block put the model's question, query and reply format inside a double-quoted _aside_exec argument, where backticks and $(...) from page or repo text would run. The agent now writes the prompt into a private .gstack/tmp file; the send block rebuilds the path from the printed basename, refuses when the file is empty, and passes "$(cat "$PROMPT_FILE")". The fixed read-only sentence stays in the shell source. aside-driver pins move from the inline prompt to the file-based send block. Co-authored-by: dgrant <dgrant@users.noreply.github.com> * fix: design briefs and approval feedback travel as agent-written files Design briefs went into $D variants/generate/check as a double-quoted argument, and approval feedback was spliced into a single-quoted echo of JSON, so a quote, backtick or $(...) in either broke the command or ran. The agent now writes each brief into a private .gstack/tmp file and the block passes --brief "$(cat "$BRIEF_FILE")" (the CLI's --brief-file expects a JSON brief, so plain text uses --brief). variants and the later check reuse the same brief file. approved.json is built with jq -n from --arg values and --rawfile feedback, so the feedback is data. The office-hours mockup flow creates its brief and feedback files in one block. design-approval-flow now writes the feedback file (with quotes, backticks and $(...) in it) instead of substituting <FEEDBACK>, and checks the file is removed after the save. Co-authored-by: dgrant <dgrant@users.noreply.github.com> * fix: carve office-hours on every host to stay under the 160,000-byte ceiling With its brief, feedback and research prompts moved into files, the inlined office-hours render is 161,329 bytes on Codex and 160,302 on Factory, over SKILL_BYTE_CEILING, so gen-skill-docs --host all fails. Per the C4 policy (carve every skill whose external render exceeds the ceiling), office-hours joins ship and plan-ceo-review in CARVED_ON_EVERY_HOST. The Claude render is unchanged; external hosts read its sections relative to the installed skill. external-host-carve and qa-lazy-sections now include office-hours in their carved-skill lists. Co-authored-by: dgrant <dgrant@users.noreply.github.com> * chore: regenerate after the free-text mktemp guard * fix(codex): prompts and review focus travel as agent-written files Co-authored-by: dgrant <dgrant@users.noreply.github.com> Challenge and consult prompts, and the /codex review focus text, are written by the agent into private .gstack/tmp files and fed to Codex on stdin or read back with cat; no heredoc carries them. Consult's new and resumed runs share one block (and one parser) so no placeholder stands in for the parser; a resumed run reads its session id from .context/codex-session-id and checks it. Adds the A3 group 2 placeholder allowlist block. * fix(benchmark-models): an inline prompt goes through a prompt file Co-authored-by: dgrant <dgrant@users.noreply.github.com> * fix(land-and-deploy): PR head repository and branch travel as data Co-authored-by: dgrant <dgrant@users.noreply.github.com> The branch check reads headRepository/headRefName from gh's JSON into quoted shell variables; ls-remote, the confirm-first delete and the local delete use those variables instead of model-substituted names. The postfail test now pins the variable form it replaces. * fix(office-hours): the reviewer receipt must match the receipt grammar before the check runs Co-authored-by: dgrant <dgrant@users.noreply.github.com> * fix(setup-gbrain): cleanup-orphans collects the PAT with read_secret_to_env Co-authored-by: dgrant <dgrant@users.noreply.github.com> * fix(ship): PR body drafts and rejected test paths are agent-written files Co-authored-by: dgrant <dgrant@users.noreply.github.com> The PR body is written into two private drafts (through ## Documentation, and from ## Test plan on); the scan block splices Step 14.5's section between them into the scanned temp file and deletes the drafts. Rejected test paths are read from a written list, and paths outside the repository are skipped. Tests that ran the old heredoc blocks now write the drafts and list first. * fix(sync-gbrain): orchestrator flags must match an allowlisted grammar Co-authored-by: dgrant <dgrant@users.noreply.github.com> * fix: the gbrain page save and external-host placeholders pass the free-text rule The gbrain save block passed the whole page through a quoted heredoc inside --content; the agent now writes it to a mktemp file and gbrain put reads it on stdin. The Claude Code outside-voice runtime root and session mode, and the feature slug, join the allowlist with grammars; unused entries are dropped, and a test keeps every allowlisted placeholder tied to a real use. Co-authored-by: dgrant <dgrant@users.noreply.github.com> * fix: skill-specific Aside research queries travel as prompt files investigate, office-hours, plan-ceo-review, review, plan-eng-review, design-consultation Phase 2, the shared test bootstrap (B2) and QA's test bootstrap each carried their own _aside_exec call with the model-filled query inside a double-quoted argument. Skills that already carry the Web research block now give only the prompt-file text and use that block to create, write and send it; design-consultation, the shared bootstrap and QA create the file and send it with the new {{ASIDE_RESEARCH_SEND}} resolver, which is the same send block the research section uses (read-only sentence stays in the shell). aside-driver's B2 pin and qa-browser-preservation's bootstrap contract move from the inline call to the prompt file and the shared send resolver. Co-authored-by: dgrant <dgrant@users.noreply.github.com> * fix: prose design commands pass briefs and feedback as data The DESIGN_SETUP command list taught `iterate --feedback "..."`, and design-consultation's late-change and plan-design-review's slop-regenerate lines showed `--feedback "<feedback>"` / `"..."`; design-shotgun's retry showed `generate --brief "<brief>"` and `evolve --brief "<brief>"`. Feedback now reads "$(cat "$FEEDBACK_FILE")" from an agent-written mktemp file (the setup note says how), and the shotgun retry reads the variant's own brief out of briefs.json with jq instead of retyping it. Co-authored-by: dgrant <dgrant@users.noreply.github.com> * test: record parity sizes after the free-text file mechanism; refresh ship goldens Measured union and skeleton sizes for ship, plan-eng-review, plan-design-review, office-hours, design-consultation, review, qa, browse and design-shotgun after every posting, prompt and brief site moved to agent-written mktemp files; the ship goldens pick up the question-tuning slug and the dropped free_text field. * test: parity caps for the combined wave head (rounded up past the measured ratio) * fix(plan-ceo-review): the mode handoff helper reminds the model to lead its chat with the line (#3038) Periodic run 37272185151's auto-decide-preserved red replays as the Step 0E handoff-line class: the model ran gstack-ceo-mode-handoff --auto, got the exact line, then opened its chat with a heading and a paraphrase, so the user never saw the auto-decide line (the terminal collapses tool output) and the detector saw no outcome. The helper now prints a one-line reminder on stderr beside the line; the detector accepts the line followed by that reminder and nothing else. The ship Greptile check test reads its template (test-of-test ratchet). * test: every troubleshooting link in shipped code resolves, relative or full URL * docs(todos): Oct 6 fix-wave follow-ups * v1.91.32.0 release notes: Codex works from zsh, free text never enters shell commands, honest failures * fix(plan-ceo-review): show the mode handoff line through a PostToolUse hook, not model compliance (#3038) The stderr reminder beside the handoff line measured 1/10 on auto-decide-preserved: the model still opened its chat with a paraphrase ('I've set review mode to HOLD SCOPE...'), so the user never saw the line and the detector saw no outcome. /plan-ceo-review now declares a Claude-only PostToolUse hook that shows the helper's exact line as a system message. Diagnostic measurement on this head: 10/10, the line visible in every trial while no trial's chat led with it. The reminder is reverted. * fix: type the handoff hook's line filter * fix: repairs from the integrated free suite - The self-locate markers looked like heredocs to the pipe-deadlock guard; they are plain comment fences now. - The spawned block keeps the canonical 'conservative non-destructive' phrase the cross-surface drift guard pins. - The free-text block's mktemp refusal says 'mktemp failed', as every /codex temp file must. - The batch transcript read reports sizes in MiB. - docs/TESTING_INTERNALS.md gets a real 'Ship measure' heading for its anchor. - The early-question harness admits /plan-ceo-review's handoff hook by its exact command (it only adds a system message); any other PostToolUse Bash hook is still refused. - Paid file count and the unconsumed-file example follow the new seeded case; ship goldens refreshed. - The CSO scanner test points Docker at a dead socket, so a machine with the pulled image does not scan. * test(cso): a host with no Docker binary is also a no-scan host The missing-image scan test failed on a Docker-less VM (docker is not installed in a trusted directory). Every accepted message is a reason the scan cannot run here; the not_assessed, helper-owned artifact assertions are unchanged. * ci: split the USERPROFILE assignment from export (SC2155) * fix(browse): Windows runtime roots read .source-path as a native path (#3026) The new windows-setup-e2e journey showed the installed browse could not start its server. setup runs under Git Bash and wrote .source-path as an MSYS path (/d/a/gstack); Node read it as \d\a\gstack on the current drive, found nothing, and fell back to the root's own bundle, which cannot import playwright. setup now writes cygpath -m, and the CLI maps an MSYS drive path for installs written by an older setup. The journey prints the daemon and startup logs when the chain fails. * fix(ship-measure): defects from the parallel #3034 plan review - A trial passes only with a JUnit testcase that executed and passed; a --case that selected nothing also exits 0. - measure runs a no-cost --list preflight that must plan the case's own file before any paid trial. - Diagnostic trials drop EVALS_CACHE_DIR/EVALS_CACHE_RUNTIME_ID, so they never reuse a stored judge pass or publish one the gate could reuse. - Gate once: when CI runs the gate on push, the push is the gate run; the skill no longer also starts eval:bg:pr (a second paid gate). - ubi-runner: down fails unless the VM is confirmed gone, and nothing sweeps other runners' VMs by default (UBI_GC_HOURS opt-in). * test(windows): the seeded-flake fixture drives a POSIX shell stub, so it stays off the Windows lane CreateProcess cannot exec the fixture's ./evals.sh shebang script (status null on windows-free-shard 6). The measure runner is covered on Windows by ship-measure.test.ts, which uses in-process fake runners. * fix: the free-text block refuses an unset temp dir instead of writing at / The rendered block ran mktemp "$_GT/..." with _GT unset in the portability test; on Linux that fails at /, but Git Bash on Windows can write to its install root, so the block went on (windows-free-shard 3). It now uses ${_GT:?}, so an unset directory always fails closed with the mktemp message. * test: pin the guarded ${_GT:?} temp-file form * test: the Greptile check's gh stub uses a portable shebang On Windows the absolute bash path has a space, so the stub never ran and every case read 'gh not logged in' (windows-free-shard 1). * test: the Greptile check runs with the native PATH and the gh stub first The PATH-only stub bin could not run its stub on Windows (every case read 'gh not logged in'). Stub cases now keep the real PATH with the stub first; only the no-gh case drops directories holding a real gh. --------- Co-authored-by: marcosdetry <marcosdetry@users.noreply.github.com> Co-authored-by: akshaykalia <akshaykalia@users.noreply.github.com> Co-authored-by: scottgigante-hubflow <scottgigante-hubflow@users.noreply.github.com> Co-authored-by: woodsonl <woodsonl@users.noreply.github.com> Co-authored-by: bezdekt <bezdekt@users.noreply.github.com> Co-authored-by: dgrant <dgrant@users.noreply.github.com> Co-authored-by: venkatareddyg <venkatareddyg@users.noreply.github.com> Co-authored-by: CharlesBerg85 <CharlesBerg85@users.noreply.github.com> Co-authored-by: VaibDeveloper <VaibDeveloper@users.noreply.github.com> Co-authored-by: Hagetino <Hagetino@users.noreply.github.com> Co-authored-by: Rekin226 <Rekin226@users.noreply.github.com> Co-authored-by: ekson73 <ekson73@users.noreply.github.com> Co-authored-by: tianxchen <tianxchen@users.noreply.github.com> Co-authored-by: BenjaminDSmithy <BenjaminDSmithy@users.noreply.github.com> Co-authored-by: ubay887 <ubay887@users.noreply.github.com> Co-authored-by: nareshh74 <nareshh74@users.noreply.github.com> Co-authored-by: syong <syong@users.noreply.github.com>	2026-10-06 23:42:19 +00:00
Garry Tan	7697fb2981	fix(cso): install Rails gems on one thread so exports stop failing (v1.91.31.0) (#3056) ... Ruby's File.umask getter sets the process umask to 0 for an instant, and RubyGems calls it for every extracted file. With bundle install --jobs 2, a directory created by the other install thread in that window came out 0777, and the prepared-tree export correctly refused it ('prepared tree contains a publicly writable directory'). The offline install now runs --jobs 1. preparedExportRejection() also accepts the helper's real output: its top-level handler prints the bare message and exits 70, which CI reported as 'unrecognized helper failure'.	2026-10-06 20:13:27 +00:00
Garry Tan	5885157a90	v1.91.29.0 fix(cso): native runtime qualification passes for Node, Bun, Python and Rails (#3044) ... * fix(cso): unblock native runtime qualification (v1.91.28.0) Fixes the qualify-native failures from main run 37345275258: - re-bind promotion evidence for staged runtimes in cold-start tests - give npm a distinct empty global config (npm 11 refuses double-loading) - export /archives and /work through tmpfs-backed local volumes (docker cp cannot read tmpfs), with exec on /work - keep the offline seed process alive - end broker/forwarder tunnels explicitly under Bun - batch gem fetch per platform - wait for run-app's source copy before Rails db:prepare and tests - fix the Python cold fixture and add bin/rails to the Rails cold fixture * fix(cso): bounded 900 s ceiling for preparation dependency commands - PREPARATION_COMMAND_TIMEOUT_MS (900 s) applies only to acquisition and offline dependency commands; everything else keeps COMMAND_TIMEOUT_MS (300 s), and both are cut to the journey deadline - prepared-tree export refusals fail with PREPARED_EXPORT_REJECTED and the helper's fixed reason or errno, never target paths, with no retry - free tests pin the ceiling, its callers, and the rejection parser * Take v1.91.29.0 (1.91.28.0 is claimed by #3032 and #3033) * cso-diff-mode accepts every correct name for the unauthenticated webhook finding	2026-10-06 10:32:19 -07:00
Garry Tan	7fca42ad8b	v1.91.12.0 v1.91.12.0: audit fix wave, ~11-minute paid eval lanes, eval reliability policy (#2999) ... * test: delete test-infrastructure dead code (G) - exit-propagation drives the runner's real strict verdict (BunTestOutputClassifier + strictTestExitCode); delete the unused shardRunLooksTruncated predicate. - delete skill-coverage-matrix registry + its gate (nothing reads it; the floor already iterates skillCensus()). - delete touchfiles-facade export-parity tests (Bun fails missing imports at link time) and the duplicated E2E_TIERS tier-value test. - delete brain-cache-spec TRANSPORT_DEFAULT_POLICY, SKILL_RUN_RETENTION_DAYS and the now-unused BrainTrustPolicy type with their literal tests. AUTOPLAN_PREFLIGHT_BUDGET_BYTES stays: skill-preflight-budget enforces it against real resolver output. - delete audit-compliance's JSDoc-comment grep. * test: replace product tests that fake the product with real-boundary tests (F) - design: serve.test.ts drove an inline mirror server; now two tests run the real serve() on an ephemeral port (reload confinement, submit exit 0). - setup-gbrain: rollback + voyage tests execute the template-extracted init blocks (3 sites) instead of drifted local bash copies. - terminal-agent: internalHandler source greps replaced by a behavioral /internal/grant + /internal/revoke auth matrix (no/wrong/valid token). - /health: server-security-surface and the server-auth / security-audit-r2 / sidebar-tabs source greps fold into one liveness-only check on the real body; the L4 sidecar wiring gets a behavioral /pty-inject-scan test. - delete tautologies (browser-manager onDisconnect, memory-command #12), ios swiftui tap fixture self-check, memory-ingest put_page grep, detach source greps, sidebar-agent absence pins, dead-CSS pins + the dead CSS, security-audit-r2 Task 1 + the test-only meta-commands re-export, duplicate generated-SKILL.md checks. - make-pdf coverage-gaps cases move into their owner test files. * test: delete tests of dead eval code (A) - A1: the retired Eng lexical oracle (evaluateEngSeedCoverage, isEngSeedDecisionAUQ), the completion-handoff detector and the retained corpus had no paid caller since v1.87.6; delete their 26 replay files, ~2.6k helper LOC and fixtures, and the dead blocks in 8 mixed files (live hasNativePlanTerminal / batching assertions stay). - A2: dead viewport approvers in autoplan-artifact-permission and their 11 replay files + fixtures; recorder/launcher cases stay. - A3: never-wired oracles and seeders (autoplan-phase-order, eng-finding-fixture, ceo-paired-fixture, design-ui-scope, plan-skill-completion, pty-current-screen, required-reads, transcript-section-logger); plan-seed-submission now decodes through the production createPtyScreen; section manifests name their actual guard. - A4: zero-reference helper exports, plus execGit and invokeAndObserve found by the reachability pass. - 52 fixtures orphaned by the deletions; touchfile and selection-table entries for every deleted path. * test: clean up the paid eval lane (B1-B4, B6, B7) - B1: delete paid files that assert nothing or cannot pass meaningfully: skill-llm-eval-spec and skill-e2e-spec-execute (test.todo), gemini-e2e (+ gemini-session-runner; no gemini CLI in CI), ship-idempotency (red since v1.63), the two opus-4-7 *-sonnet overlay wrappers, conductor-prose (+ its source-evaluation replay), codex-e2e-plan-format; drop their keys, scripts and census rows. - B2: skill-llm-eval grades browse/sections/command-list.md with one union judge that also carries the baseline score pin; regression-vs-baseline deleted (paid run: pass, c4/c4/a4). - B3: memory-pipeline, ios-qa, ios-qa-swift-build and plan-tune-cathedral make no model calls; renamed out of the paid glob so they run on every PR. Swift builds need GSTACK_TEST_SWIFT=1; device stub deleted. - B4: codex-e2e*, outside-voice, aside and ios-device cannot run in the CI image; excluded from the weekly lane with a tracked re-entry condition. - B6: fold opus-47's negative routing controls into skill-routing-e2e journey-negatives (paid run: 3/3 unrouted) and delete the file. - B7: delete the never-green brain-privacy-gate eval; a free gstack-skill-start test now proves consent precedes artifacts egress. * test: retire the finding-count cluster and trim its helpers (C) - C0/C1: the five never-green evals (skill-e2e-autoplan-chain and skill-e2e-plan-{ceo,eng,design,devex}-finding-count) failed on harness and budget, never on skill behavior; delete them, their touchfile/tier ids, AUTOPLAN_CHAIN_BUDGET and the dedicated eighth periodic slice (--slices 7). - C2: delete the helper groups whose only paid consumers were those files (11 modules), trim claude-pty-runner and eng-seeded-coverage to the paid closure, and delete the free replay tests whose assertions exercised only that dead code (89 files, 135 orphaned fixtures). Blocks that used dead code only as input for a live subject keep their assertions: the multiSelect default moved to plan-review-decisions, runner PTY tests use inline caller policies, and the timer-safe budget checks moved to eng-finding-retry-budget. - The eight production-touching files stay except ceo-current-decision-record (its template read only feeds the retired counter). - CARVE_GUARDS.autoplan is behavioral 'none'; TODOS records the lost chain and per-finding cadence coverage with their re-entry tests. * test: fold per-incident replay series into their detector owners (D) Twelve detector families move into one owner test each: 73 incident files become describe blocks in ceo-section-loading-fixture (stale-fill race), model-overlays, coverage-audit-evidence, autoplan-phase-observer, native-auto-decide, outside-voice-evidence, eng-first-review, plan-count-completion, plan-count-file-permission, ceo-mode-option, plan-scope-selection and plan-count-prerequisite. Each block keeps its original code and fixture, so every case still runs; only tests asserting the incident file's own touchfile registration are dropped (41). Touchfile lists that named an incident now name its owner. * test: start the plan-count history PTY on its readiness marker (H) The fake CLI prints a startup marker and the runner waits for it instead of the fixed 8 s startup sleep (8.6 s -> 0.9 s locally). eng-semantic-terminal's sleeping registration cases went with C; plan-count-timeout keeps the fixed wait because it asserts deadline behavior. * test: derive paid touchfiles from each eval's static closure (E) touchfiles.test.ts now checks, per key, that the paid file's static test/helpers and test/fixtures closure (plus fixture paths it names in string literals) is covered, and names the file, path, chain and key to fix when it is not. Free *.test.ts files are no longer touchfiles, so editing a free replay test stops selecting paid evals: 950 entries removed, 653 real closure paths added. The hand-copied inventories go: periodic-fixture-selection, fake-impeccable-touchfiles and 45 per-file selection examples. Selection for the sample edits (plan-eng-review template, claude-pty-runner, plan-count-fixture, gstack-config) loses no case under either profile. CONTRIBUTING documents the rule and its lower bound. * test: skip hollow tier shards and census judges in the paid planner (B5) A paid file is now skipped for a tier lane only when every E2E id it registers is known statically and none has that tier; ids come from the touchfile registrations and literal testName/*IfSelected arguments, so a comment or skill path that quotes another id cannot unschedule it, and computed names keep today's scheduling. --list and the manifest show each skip as "skipped: no E2E_TIERS id has tier <tier>". The weekly gate census drops the LLM judges (--skip-judges); they still run in the periodic census and PR gate lanes. Gate lane 52 -> 42 files, census 41; periodic 77 -> 69. * test: run seven paid evals on the current default capture model (B8) skill-e2e-{auq-matrix,plan-format,qa-bugs,retro,workflow} pinned claude-opus-4-7 and skill-e2e-office-hours plus -brain-writeback pinned claude-sonnet-4-6; none tests a historical model, so they now capture with resolveEvalModel('capture'), and the free harness tests that execute these registrations receive the same resolver. The paid re-pin run passed all of them. skill-e2e-{design,office-hours-phase4,plan-prosons,plan} keep claude-opus-4-7: six of their cases failed on the default model (three timeouts, a missing report file, a format miss and a posture score of 3), so per the plan's fallback they keep their pins with a TODOS entry. The pre-spend estimate and drop threshold are in docs/test-audit-2026-09.md. * test: guard the reduced suite against new test-of-test files - test/test-of-test-ratchet.test.ts records the 228 free tests that import only test/ code and fails on a new one, naming the owner test to extend instead; a stale baseline entry fails with the remove instruction. - test/helpers/resolve-repo-path.ts is the one specifier/literal resolver for the ratchet and the touchfile closure invariant, with its own unit tests. - CONTRIBUTING "Test tiers" describes the paid-failure workflow (fix, then one row in the detector's owner test) and the ratchet; TEST_PORTFOLIO gains the detector -> owner-test table and no longer claims an Autoplan chain eval. - TODOS: automatic exclusion policy for chronically red periodic files (P3), the deferred native-completion table collapse, the unused CEO payment seeder; the PTY readiness item is narrowed to the paid runner. - docs/test-audit-2026-09.md collects the triage, security mapping, inventories, selection proof, behavior-commit decisions and retained false positives. * v1.91.8.0 test: smaller suite, derived paid selection, retired never-green evals Release metadata for the test-reduction branch: VERSION 1.91.8.0 (1.91.7.0 is claimed by #2983), CHANGELOG with the measured before/after table and a contributor section, durations re-recorded on Ubicloud standard-16 (857 files, 0 failures), the agents digest, CONTRIBUTING's after-measurement row, the B8 fallback TODOS entry, and the after metrics, kept-vs-plan notes, B8 run and census estimate in docs/test-audit-2026-09.md. * fix(ubicloud): skip retrieval globs that match nothing instead of reporting a failed pull * test: pin DISABLE_AUTOUPDATER in hermetic env and capture corrupt-seed warning Both EVALS_HERMETIC branches of buildHermeticEnv now carry DISABLE_AUTOUPDATER=1 (the allowlist scrubbed the workflow's copy, so every PTY screen showed the updater's npm-prefix failure). Per-test overrides still win. The corrupt durations-seed test now captures its expected warning and restores the console spy. * style(cso): format lib/cso TypeScript with pinned Prettier Mechanical reformat only. Minified transpile output is byte-identical for 21 of 22 files; witness.ts differs only in three regex flag orders (/mi -> /im), which JavaScript canonicalizes. Source-text assertions over lib/cso now compare whitespace-insensitively with the same tokens. * fix(cso): import join for compiled-launcher assertion witnesses Compiled installs always take the non-Bun branch, which called an unimported join and threw before any runtime-tested assertion could be witnessed. The child command selection is now a pure, platform-aware function; a missing sibling launcher fails with its expected path. * fix(browse): make connect --supervise actually respawn a crashed server The supervisor respawned with a block-scoped env that no longer existed, so every attempt threw and the loop gave up after five tries. The headed env is now one pure helper used by connect and respawn, the loop is an injectable runHeadedSupervisor with behavioral tests, failures name the daemon log and relaunch command, and connect's usage advertises --supervise. * test: one finite PR world for the shared-libs fixture; name dual-voice probe evidence The shared-libs shim served 2 PRs for pulls?state=all and endless full pages for state=open. gh pr list, pulls?state=open|all|closed (per_page/page, short last page, direction) and search/issues now page one deterministic table: PR 7, 600 older open PRs, PR 42 and 3 closed PRs, so five 100-item open-metadata pages still leave older open PRs unchecked. The Contents API lists pinned directories (the captured attempt got 404 for contents/ and contents/src while files resolved, then fell back to a raw host), unknown endpoints return 404 instead of repo metadata, and the read-only detector is unchanged. Free tests cover view agreement, the budget bound, gh/curl agreement and the empty world. Dual-voice outside-voice failures now report probeToolUseId, probeMode and the canonical-match result with the reason the probe output was rejected. * feat: require a zero-error product typecheck and a test type-debt ratchet Adds tsconfig.json (strict) over product code, fixes its remaining 90 diagnostics (type-only, interface corrections, and explicit narrowing), and adds a typecheck job to the required free-tests aggregate running bun run typecheck, the test-code ratchet (identity -> count baseline, fails on new, repeated, or unlocked fixed diagnostics), and the lib/cso format check. Reuses fixes from #2447 where they still applied. * test: follow the headed env helper and the typecheck gate in source-shape checks * fix(test): pin the package.json change kind in shared-input selection tests computePaidCaseSelection read the version-only exemption from git even when changed files were injected, so the shared-input test failed on main and on version-only branches. The exemption is now an optional input; the test pins a real package.json change and covers the version-only case. * test: judge plan-count completion on structured evidence, not wording Replaying run 36385945043's two Design attempts showed the existing routes rejected correct endings: attempt 1 at the typed-completion path field ('- Reviewed plan written to …' is not a 'Plan written to' line), attempt 2 at the leading-fence veto (its final message opens with the dashboard). nativePlanTerminalPreconditions is the structural prefix of hasNativePlanTerminal (behavior unchanged). structuredPlanCompletion adds, inside the existing nativeSummary branch: a complete report (Design binding for Design), a completed review-log row for the expected skill appended during this attempt under the child's GSTACK_HOME/project slug (resolved with bin/gstack-slug) and stamped with the fixture commit, timed between the report/last answer (second resolution) and the final native message, a final message with stop_reason end_turn (now carried on public transcript messages), and no visible question or permission prompt. Timeout summaries add idleFor and lastTerminalCandidate. Terminal and throw captures copy the plan file and review-log rows into the artifact directory; copies are best-effort and recorded in evidence-copy.json. Free regressions: both captured Design endings (trimmed fixture with provenance; report, row and end_turn reconstructed and labelled), the negative controls, and real-PTY completion/timeout runs through the real review logger. * test: structural Design count boundary; TODO proposals are not findings Replaying run 36385945043 through the Design count predicates: routing, focus and learnings setup was not recognized as setup, Issue 1 was counted pre-review in both attempts (the boundary fired on it), and attempt 2 counted the Font TODO proposal as a finding (review=4 and review=5 for five issues). The paid caller now starts review at the first answered native decision that is not setup (recognized packet, or setup header/question ID), a completion handoff, artifact rendering or a TODO proposal (the review's Add to TODOS.md / Skip / Build it now menu). TODO proposals are recorded as administrative extra decisions. The replay asserts each counted call: both attempts review=5 (Issues 1-5). isDesignCountFirstReview and its controls are unchanged. * test: CEO classifier throws name the question and matched predicates Replaying run 36385945043's FAN-1 and ERR-1 throws (ledger rows reconstructed from rendered diffs) through ceoPaymentFinding: the email obligation's row, subject, option and proposal predicates pass and the ELI10 explanation-defect predicate fails first ('lets that exception fly out', 'the error bubbles up'). Binding the defect to the named ledger row instead (the planned fix) was tried and reverted: scoped to the email seed it flips 30+ existing cf74 still-rejects replays, which require a vocabulary-free, ledger-bound email question to earn credit only through a complete saved comparison. With FAN-1's rendered currentDecision payload reconstructed, the recorded- decision path counts it, so the real saved plan (not uploaded) must have differed; failure artifacts now retain it. The classifier stays fail-closed and unchanged. Its throw now prints the header, the first 200 question characters and each obligation's predicate results. Free regressions with provenance and negative controls: an unrelated question, an email question whose row says it is already rescued, and a ledger ID whose row belongs to another seed. * chore: regenerate the test type-debt baseline on top of #2994 * fix(typecheck): strip the checkout root from ratchet diagnostic identities * fix(test): recognize ledger row-ID split candidates so collection stops at the last ACK Run 36385945043's split-overflow case asked all five candidate decisions by 8m55s, but the live candidate check required the question to open with "E1:" and every option to be a known disposition. The skill cited ledger row IDs ("D2.1 — R-E1: …") and offered "Hold, discuss first", so no candidate was recognized and the attempt ran the whole review (1302s). Identity now comes from the native header; the question must open with that candidate's ledger reference, name only that candidate, and offer exactly one include, defer and cut disposition. The selected answer must still be one of those three. The semantic evaluator and every existing negative control are unchanged; a trimmed capture from the run adds the positive case and four row-ID negative controls. * fix(test): stop the eng batching eval once its floor is proven The case's only verdict is reviewCount >= FLOOR (3). Run 36385945043 had three distinct acknowledged review decisions at 6m41s but kept answering until the ceiling (7) at 12m13s. The registration now passes the runner's existing isCollectionComplete stop once FLOOR non-setup, non-administrative review decisions are acknowledged; the floor check, ceiling, budget and counter are unchanged. A child-process registration test proves the stop predicate and that below-floor and timeout outcomes still fail. * test: add the non-blocking 'marathon' E2E tier Full start-to-finish flows move out of the blocking lanes. E2E_TIERS and E2ETier gain 'marathon'; describeE2ETier('marathon') is enabled only when EVALS_TIER=marathon, so the gate/PR and periodic lanes (and the gate census) never run those cases. The PR profile accepts marathon ids as scheduled elsewhere and defers them with their own reason, even on full fallback. * test: move the full office-hours workflow to marathon; add a periodic design-draft checkpoint The full startup workflow runs 1–3 real spec-review rounds (~280s each) and hit its 1200s capture in run 36385945043 at finalize. Review depth is the product's loop, so the case cannot fit a blocking lane without cutting rounds. It is now marathon tier with every assertion unchanged. skill-e2e-office-hours-design-draft.test.ts (periodic) runs the same fixed interview only through the Write that creates the design (269s in that run) and applies the full validator's design-draft checks, the required section reads and the launch/foreign-skill-read guards. validateOfficeHoursDesignDraft is extracted from validateOfficeHoursCompletion, which still applies it. Selection: office-hours-design-draft is registered periodic; the marathon-only file is already excluded from the gate and periodic plans by the B5 planner rule. Tier-alignment regexes and the valid-tier check accept 'marathon'. A type-only cast in plan-scope-selection.test.ts removes a diagnostic whose union print order made the ratchet identity unstable; baseline tightened. * test: supply the split-overflow fixture's HOLD SCOPE mode as a prerequisite The split actor always answered 0E's mode question with HOLD SCOPE. The skill skips that question on an explicit choice, so the fixture now states it and the attempt starts at the five candidate decisions (about 1.5 min earlier in run 36385945043). Candidates, actor policy, floor and semantic evaluation are unchanged; the fixture test pins the supplied choice. * test: start the eng batching eval with its setup prerequisites supplied Routing setup and cross-project learnings (D1/D2 in run 36385945043) are never counted and are not what the case measures. The registration now uses the runner's existing preconfiguredReviewActor so the attempt starts at the review; engSetupAUQ still vetoes any late setup question. The registration test pins the option. * test: count the design-draft paid file and defer marathon ids in PR selection pins The discovered paid-file census grows by one (skill-e2e-office-hours-design-draft). Full-fallback PR selection defers every non-gate id; the shared-input pins now expect periodic and marathon ids there. * fix(review): resolve the judged revalidation, setup-authority, plan-gate and findings-record ambiguities The census review workflow judge scored clarity/actionability 3 on both attempts: smoke-clock limits appeared to forbid post-repair revalidation, the caller deadline was undefined, 'ask for setup' conflicted with the report-only browser rule, fallback-sourced HIGH discrepancies had no gate decision, and the Step 5.8 record omitted adversarial findings. * fix(office-hours): load the builder section for every builder-mode reply Both census builder-wildness attempts answered a direct request for adjacent unlocks without reading phase-2b-builder-brainstorm.md, whose trigger read as applying only to the generative questions. * fix(sync-gbrain): define Step 4 helper args and one atomic write path Both census read-ready attempts spent turns reading the helper source to resolve <user-args>, inspecting fixture internals kept inside the repo, and reconciling 'Read + Edit' with the tmp+mv atomic write, then hit max turns before the verdict. * refactor(evals): share the import-closure walker and add the E2E shard reuse identity sourceDependencyClosure moves from the workflow-judge adapter into scripts/eval-input-cache.ts unchanged, so judge keys stay byte-identical. scripts/e2e-shard-reuse.ts builds the consumed-input identity of one PR-lane E2E shard (test import closure, every registered case's touchfiles, globals, runner/workflow/setup actions, child env pins, CI image, Claude CLI) and fails closed on anything unknown. Marathon joins the always-fresh purposes. * feat(evals): ~12-minute blocking paid lanes and a non-blocking marathon lane - Planner budget mode (--slice-budget S --jobs J): recorded per-tier wall times pack into as many ~9-minute executors as the work needs; the plan records per-slice estimates and the CI job timeout (supervised worst case + 20 min). evals.yml and evals-periodic.yml derive matrix size and timeout-minutes from it; max-parallel covers every slice at once. - Case shards: plan/design/review-army/shared-libs(-paths) run one registered case per process (<file>#<case id>, exact name pattern, exactly one case). - Retry rule: a timed-out attempt is a verdict. Only files whose every case budget is CAPTURE tier or shorter keep one retry; walls shrink to match. - Marathon tier: positive selection, excluded from gate/periodic planners, run by the new evals-marathon.yml (weekly + dispatch, fresh, own report). - PR-lane E2E reuse of verified first-attempt passes on identical inputs; the report rejects reuse outside the fast PR profile. - Duration seed from census run 36385945043, per tier and per case shard. * docs: blocking lane budget, marathon lane, retry policy and E2E reuse * chore(typecheck): lock in two fixed test diagnostics * fix(ci): drop a duplicated env/jobs block in evals-marathon.yml * test(ship-docsync): shard the doc-sync lifecycle by case and drop the duplicate dispatch-only case ship-docsync ran the same fixture and prompt as ship-docsync-completion and asserted a subset of it. The file now runs one case per process, so its lane wall is its longest case instead of half the sum of thirteen. * fix(evals): plan CI-unrunnable cases as excluded entries, not empty case shards design-review-fix drives the Aside browser and registers test.skip on Linux runners, so its case shard executed zero cases and failed the exact-one-case check in proof census 36597762183 (eval-slices 6). CASE_CI_EXCLUDE (reason + tracking, beside PERIODIC_CI_EXCLUDE) now turns such cases into excluded manifest entries that --list and the manifest surface; every planned case shard still must execute exactly its case. * docs(todos): list the case-level Aside exclusion with the CI-unrunnable evals * fix(plan-ceo-review): restore experience-first expansion framing, require the mode handoff, skip pacing menus Census 36597762183: both mode-routing runs logged provenance and moved on without the mandated handoff chat; the EXPANSION run asked an unauthorized batch/narrow pacing menu instead of the first per-addition question; the expansion-energy proposals led with the spec because v1.87.6.0 dropped 'lead with the felt experience'. The HOLD review detector also rejected a decision whose grounding line named no plan file although the owned source Read binds it. * test(outside-plan-disabled): bind quoted prior-record values by their sentence, not phrase order The parent obeyed the off switch and twice named the seeded completed record as pre-existing, once with the quotation after its owner and once with slash separators; the order-specific stripper counted both as current completion. Timestamp, location, current-claim and value-match controls still reject. * test(outside-plan-disabled): compare named record timestamps as instants; negated authorship is not a current claim The repair rerun named the seeded record by its ISO second (2026-09-29T16:58:52Z vs .727Z) and said 'I did not write'; both were misread as a foreign timestamp and a current write. * test(ceo-section-loading): recognize an arrow-ordered stale-fill execution by event roles The census review traced the seeded race as 'R1 miss -> R1 store read (v1) -> W commit v2 -> W cache.delete -> W fulfills -> R1 cache.set(v1) -> R2 (begun after W) hits v1', but the in-flight gate only accepted race vocabulary or fixed sentence shapes. Order, actor, version and dismissal mutations still fail. * test(design-floor): answer the seed-declared all-seven 0D focus menu while it is pending The actor declares 'Design: review all seven dimensions', but its picker reused designReviewSetupAUQ, which only matches already-answered calls (and a narrower header/label set), so the pending D1 focus menu was never answered and the case waited out its 609 s deadline. The skill's Step 0D requires asking; the fixture now answers it. * test(ceo-mode-routing): accept the skill-mandated Note form and Recommendation reason as HOLD posture HOLD Defer/Keep briefs must use 'Note: options differ in kind' (preamble), but the answered-HOLD path demanded a Completeness score, rejected a one-line Net with a semicolon, and read posture only from ELI10. The rerun's brief applied HOLD SCOPE in its Recommendation reason. Revert the ineffective 'always'/'handoff chat' wording: two runs still skipped the mode handoff. * test(qa-bugs): keep claude-opus-4-7 after qa-b6-static stalled on the default model qa-b6-static timed out on claude-fable-5-1 in census 36597762183 and in one of two targeted reruns. Both times the stream stopped mid-message with no pending tool, right after the model found the disabled submit button, and stayed silent until the 300 s deadline. Per the B8 fallback, re-pin with a TODOS entry; budgets and retries are unchanged. A rerun on opus-4-7 passed (125 s, 5/5 detected). * test(evals): add E2E_KINDS, BEHAVIOR_WHY, EVAL_POLICY and CASE_QUARANTINE skeletons Every E2E_TIERS and LLM_JUDGE_TOUCHFILES key starts as 'rule'; BEHAVIOR_WHY and CASE_QUARANTINE start empty. EVAL_POLICY pre-registers the approved panel (3, majority 2), quarantine entry 0.95/10 and exit 0.97/10, 10% cap, 8-weekly-run expiry, Fisher drift alarm and one INFRA re-dispatch. * test(evals): add trial records, panelVerdict, expectContract and trial-outcomes JSONL EvalTestEntry gains case_id, kind, trial, panel, failure_class and policy_version, stamped from the runner's TRIAL_ENV on isolated trial shards. panelVerdict() is the single verdict function (INCOMPLETE on missing or duplicate trials, contract veto at any count, quarantine hard-break rule, INFRA/INCOMPLETE machine classification). expectContract() records failure_class 'contract' on the collector entry and a sidecar before throwing. trial-outcomes JSONL has a fail-closed writer and a data-only reader. * test(evals): pin the fail-closed rule-shard gate through the real --report path Synthetic slice artifacts for rule fail, timeout, missing slice, unreported entry, hollow, never-started, collector failure and wrong-slice reports all exit red before the panel-verdict gate change lands. * test(evals): retire every paid automatic retry Paid evals never retry (approved 2026-09-29): delete SHORT_CASE_RETRY_FILES and retriesWithinCaseCap, drop the retry fields from the registered wall rows (walls now cover one run plus reserve), make retriesForFiles return 0, pass --retry 0 explicitly, and drop --retry 1 from the package.json paid scripts. Add the eval:pass-rates alias. Tests that pinned the old retry allowance are updated as a policy change; review-finalization-budget now proves late-result recording under the production zero-retry arguments. * test(llm-judge): sample every judge as a pre-registered 3-sample panel Each of the 24 skill-llm-eval judges now draws EVAL_POLICY.judge.samples independent samples of the same prompt concurrently inside the unchanged JUDGE_MS budget. Numeric dimensions gate on the per-dimension panel mean against the unchanged threshold; booleans (would_browse, consistent) on a strict majority. An erroring sample fails the whole panel and is never resampled; a refusal is an unscored panel only when every sample refused. callJudge's 429 backoff stays: it is transport before any model output. The workflow-judge cache stores and validates only complete panels, and its identity now records the panel and zero file retries. Harness tests that pinned one provider call per case now pin the panel size. * test(evals): classify every live case and re-select a case when its kind changes E2E_KINDS: rule by default (191 E2E ids), 22 behavior cases whose verdict is a live model choice with an acceptable sub-100% per-trial rate, each with a BEHAVIOR_WHY tolerance, and 25 judge entries (the 24 workflow judges plus the fixed-fixture llm-judge-recommendation rubric check). Contract-shaped cases (ask-before-decide, plan-mode no-writes, mandated steps, secrets, the batching floor) stay rule. Behavior requires a known literal registration and an exact Bun test name so the case runs as its own trial shard. Map-diff selection now diffs E2E_KINDS and BEHAVIOR_WHY per key, and a base revision without them selects every key, so a kind flip runs the panel it introduces. test/eval-kinds.test.ts enforces coverage, tolerances, isolatability and the reviewed counts, printing the literal to add. * feat(evals): per-case pass rates with Wilson intervals, identity series and quarantine policy scripts/eval-flake-rank.ts becomes eval:pass-rates (eval:flake-rank stays an alias, and the legacy aggregate stays exported). It reads eval-store's trial-outcomes JSONL from the last N completed evals-periodic runs on this branch and main (gh, downloading only the trial-outcomes artifact, cached and size-capped, parsed as data), plus local eval dirs, and prints per-case per-trial pass rates with 95% Wilson intervals. A series is a case's own touchfiles minus GLOBAL_TOUCHFILES (caseSeriesIdentities, for the report job to stamp), per model, CLI version and policy version. Labels: INCONCLUSIVE, BROKEN, FLAKY, FAILING, PASSING. --backfill imports legacy slice artifacts as pre-policy trials (first attempt only, attributed by registry id, never guessed) for display only. --gate fails with ACTION REQUIRED on post-policy evidence only: drift below the quarantine entry rule, a rule case behaving like behavior, a one-sided Fisher drop against the previous identity (Holm-controlled), and quarantine entries that met their exit rule, expired after 8 weekly runs, broke the 10% tier cap or are invalid. CASE_QUARANTINE entries now carry a failureClass (detector, harness or model-latency); a product defect has no class and is never quarantined. The policy test pins EVAL_POLICY's approved constants. * feat(eval-pass-rates): attribute legacy records by the exact slug of their display name * ci(image): pin Claude Code 2.1.284 so the eval model is recognized 2.1.251 logs [claude-code:unrecognized_model] for claude-fable-5-1, the eval capture/judge default. 2.1.284 does not. The gate PTY smoke subset (plan-ceo/plan-devex plan-mode, plan-mode-no-op) parses on the new TUI; plan-design-review-plan-mode passed at 293 s on 2.1.284 and timed out at 300 s on 2.1.251 on the same tree. * test(eng-batching): grade the floor once the review report is complete A completed GSTACK REVIEW REPORT ends the review, so the review-question count is final there. Run 36606688266 wrote its report at 1,248 s and closed the session at 1,318 s; the case now stops collection and applies the unchanged floor at the report instead of waiting out the session. No budget changes. * test(eng-batching): bind unsourced native briefs through the report's target Run 36606688266 asked ten separate native review questions (D1-D9 bound to ledger records R1-R9) and failed reviewCount=0 < FLOOR=3: its briefs named the plan by title instead of citing PLAN.md, its report declared 'Review target (fixed): PLAN.md' under '# Engineering review: <plan>', and it kept an unfenced copy of the plan's own H1. The named-source route now accepts those spellings and non-inline ledger briefs. The same replay rejects a foreign, mixed, duplicate or missing target, another plan's title or copied H1, a brief naming another plan or file, a mismatched saved brief, and re-asks. The run-36597762183 capture still counts 3. * fix(plan-design-review): treat a designer with no API key as unavailable Both proof runs (36597762183, 36606688266) printed DESIGN_READY, hit 'No OpenAI API key found' on the first $D variants call, then hand-built HTML/CSS wireframes, screenshots and a comparison board for ~195-245 s before the first review question; the second run timed out at 600 s. A failed first generation now takes the existing text-only path, and the skill forbids substituting hand-built mockups. * fix(deslop-shared-libs): read related sources together within the turn limit Run 36606688266's opportunity audit read sixteen sources one per turn and stopped at error_max_turns; the passing run 36597762183 read the same files in three batched commands. The skill now says turns are bounded and asks for parallel reads or one read-only command per step. * test(ceo-mode-routing): submit a mode review that scrolled past the viewport Run 36606688266 bundled routing, learnings and the mode choice into one native call. Its review panel was taller than the terminal, so the tab bar scrolled off, ceoModeSubmissionInput returned null for 240 s and HOLD SCOPE was never submitted ('no posture match'). With no bar on screen the viewport must still end at the focused Submit prompt, and the accumulated screen text supplies the one complete panel, authenticated exactly as before. Replay controls reject another mode, an unoffered answer, an altered question, a quoted panel, trailing output, a moved cursor and an answered or changed call. * docs(evals): document the pre-registered verdict policy, quarantine, pass-rate history and arithmetic AGENTS.md replaces the retry rule with the approved policy text (no retries; kind fixes trials; no added trials, samples or dispatches after a result; quarantine by CASE_QUARANTINE only; one INFRA/INCOMPLETE re-dispatch) and notes that a pre-registered fixed panel is not rejudging. CONTRIBUTING gains the kind rules, the judge panel, eval:pass-rates and an 'Add a paid eval' checklist. TESTING_INTERNALS describes verdicts, quarantine, history and the arithmetic, including the rule term: 1 trial vs 2-of-3 red rates at p = 0.99/0.95/0.90/0.70/0.30 and lane all-green probabilities for the current 191 rule / 22 behavior / 25 judge registry. * feat(evals): trial planner, slice exit split and panel-verdict report Planner: behavior and quarantined cases become panels of isolated trial shards (<file>#<id>~t<N>) bound by EVALS_SELECTION_JSON=[id] and the exact test name; the file shard excludes them by name. Trials of one case never share a slice, result slugs are unique, panels are validated whole, unknown registrations throw, and the planner prints a capacity preflight. Executor: each trial shard gets its TRIAL_ENV identity and a trial record (outcome, failure class, cause, cost); every shard writes a JUnit report. The slice exit now means execution completeness: a failed rule shard or a trial without a record reds the runner, a failed trial does not. Report: panelVerdict() decides every panel of the first run attempt (later attempts are reported, never replacing it); rule shards keep the unchanged fail-closed checks; collector records all count (no last-attempt wins); census runs enforce the quarantine cap and expiry. It writes collector-outcomes v2, trial-outcomes.jsonl (trials plus JUnit rule/judge cases), report-summary.md, and one headline + failure block with rerun commands, and flags INFRA/INCOMPLETE-only reds for the one re-dispatch. The fail-open suite gains the panel cases: behavior 1/3 red, 2/3 green with its failed trial shown, missing trial INCOMPLETE, contract at 2/3 red, quarantined 1/3 green, 0/3 and contract red, missing slice red, and a later attempt never replacing the first. * chore(evals): refresh paid duration seeds from proof runs 36597762183 and 36606688266 Both tiers, merged in run order (the later run wins). Notable: split-overflow 1332s -> 504s, section-loading 604s -> 342s, mode-routing 575s -> 444s; multi-finding-batching 734s -> 1318s (its red path in run 36606688266). * feat(evals): stamp trial series identities and fit panels to the live registry - scripts/eval-trial-series.ts stamps series_identity (eval-flake-rank's caseSeriesIdentities) on a report's trial-outcomes JSONL as its own step, keeping the history tool out of the paid runner's closure; TrialOutcomeRecord gains the optional series_identity field. - Slice-count plans let a registered trial spill into an ordinary lane when its siblings hold every long lane, so panels never share a runner. - Re-audited test-selection.ts (Stream B added the E2E_KINDS/BEHAVIOR_WHY map-diff; no new module loading) and repinned its hash. - Detach and release floors now count trial shards (66 periodic trials in 22 panels): periodic floor 33,821s, still under eval:bg:periodic's 67,380s. - Coordination fixtures supply the executor's trial records. * ci(evals): attempt-scoped artifacts, verdict-v2 PR comment, weekly pass-rate gate and one INFRA re-dispatch - Slice, census and marathon artifacts carry -a<run_attempt>; reports download them per artifact (no merge), so records never overwrite and a re-run never replaces the first attempt's verdict. - Planners pass --max-parallel for the capacity preflight (24/16 unchanged: the refreshed periodic plan needs 24 slices, the gate census 12). - PR comment: jq-only job reads collector-outcomes v2 (headline, sanitized failure block); the group_by(.name)|last recomputation is gone. - Reports stamp series identities, upload trial-outcomes-* for history, and shard logs upload always (a failed trial no longer reds its runner). - Weekly report: headline + failure block of both lanes in the issue body, the eval:pass-rates --gate step (fails closed without history), close the issue on a green run, and UC-E1: when every red is machine-classified INFRA/INCOMPLETE, one re-dispatch as a new run in its own concurrency group (redispatch_of), both runs reported. * feat(evals): planner-side whole-panel reuse and negative receipts The planner job restores this PR's receipt store once and ships a single filtered set with the plan: a pass or panel receipt with a same-or-newer FAIL for its input identity is dropped, and a panel receipt ships only as a whole PASS panel (re-verified with panelVerdict) from one run. Executors read only that set (no per-slice cache restore or save), so every trial of a panel sees the same receipts; a trial reuses its own record from the panel receipt, keeping a split PASS's failed trial. Trial identities drop the trial index (run-scoped) and bind the panel policy. Executed shards carry their input identity; the report turns a whole fresh PASS panel into a panel receipt and a FAIL panel or failed rule shard into a negative receipt, and marks a panel that mixes reused and fresh trials INCOMPLETE. The report job merges plan, slice and report receipts (newest per file) and saves one store per run. Also fixes two TS2352 casts in browse/test/dia-macos-qualification.test.ts whose diagnostic text drifted with program order (baseline locked, fix only). * feat(evals): --case/--trials local diagnosis and panels in local sharded runs bun run scripts/test-paid-shards.ts --case <id> [--trials N] runs N independent trials of one case through the CI panel runner (trial shards, TRIAL_ENV identity, name-pattern isolation) and prints its panelVerdict(); N defaults to the case's policy panel and CI never reads it. The local sharded path (test:gate:sharded, test:periodic:sharded) now plans the same trial shards and exclusions as CI and exits on execution completeness plus panel verdicts. * test(pty): grant an owned Create pane whose title row is cropped The targeted batching rerun on Claude Code 2.1.284 left its first report Write unanswered for 1,372 s and timed out: the viewport began at the pane's relative file row and rule, with the 'Create file' title cropped above, so the preview parser rejected the file row as foreign. That row must now resolve to the owned path and is skipped before the unchanged line-by-line preview match. Replay controls reject another file, another directory and an edited preview row. * fix(evals): tsx-safe generics in eval-flake-rank, legacy artifact names, no-retry wall docs * test(evals): record the read-only and detector-row invariants as contracts shared-libs-opportunity-judgment and review-design-lite are behavior cases: their recommendation and checklist judgments may vary, but the read-only invariant (commands, provider requests, fixture bytes, hooks, state) and the deterministic fake-engine detector rows are contracts. Both now go through expectContract, so any failure vetoes the panel. * test(judges): sample the recommendation rubric as a panel; never re-ask armJudge llm-judge-recommendation is a judge case: each fixture now draws a 3-sample judgePanel, gates reason_substance on the panel mean and the present/commits/has_because checks on a 2-of-3 majority, thresholds unchanged. armJudge no longer re-asks on a malformed verdict; it is a failed sample, as the judge policy requires. * test(evals): record a pre-turn API or CLI failure as infra recordE2E sets failure_class 'infra' on a failed session whose runner reports error_api, timeout_startup, error_output_stream or a non-zero CLI exit with zero turns and no assistant event. A model refusal, a timeout after model work, max turns, or an explicit caller pass/class keeps its ordinary classification. * test: pin every-record outcome counts and the twelve doc-sync callbacks * test(eng-batching): read the report target as a field, not a spelling The next targeted rerun (Claude Code 2.1.284) again asked eleven separate native questions and again counted zero: its briefs named no plan and its report declared '- **Review target (fixed):** `/abs/PLAN.md`' under '# Eng Review — PLAN.md: <plan>'. An unsourced brief now inherits the one current target field that names a PLAN.md file, whatever its list or emphasis markup; its ledger record still supplies the cited finding and must reproduce the brief exactly. A brief that names its plan must still match the report title. Replays of all three captures count 9, 9 and 3; controls reject a foreign, duplicate or missing target and an archived title. * fix(evals): --case list mode and name precheck; case-shard qa-callers; refresh batching and design-with-ui seeds * chore(release): v1.91.9.0 * test: settle the post-response composer before seeding; give the TPA recorder adapter its infra helper submitPlanSeed accepted a stale empty composer when the transcript recorded end_turn before the CLI repainted (late-repaint-typed-current fails 5/5 on the old helper, passes 5/5 now). The TPA recording fixture extracted recordE2E without isPreTurnInfraFailure, so every failed case threw before recording. * test(autoplan-dual-voice): unwrap Claude Code 2.1.284 subagent hand-back frames; accept read-only probe diagnostics; record before asserting Census run 36626737820: the native CEO report arrived framed and indented, so its INPUT line never matched, and the model's exact probe plus two variable echoes was not canonical. A column-zero line inside a frame, command substitution, backticks, redirects, assignments, CODEX_MODE echoes and output line-count mismatches stay rejected. The failure now records before asserting. * ci(image): keep Claude Code 2.1.251; test(ceo-mode-routing): keep HOLD's own deferrals in scope before assessing its rigor decision 2.1.284 enables per-turn effort for claude-fable-5-1: in gate census 36626737820, 66 of 84 sessions ran longer than on 2.1.251 (+20% session time, +32% thinking tokens) and 11 cases timed out on unchanged budgets. HOLD SCOPE's 0G step asks its own defer/keep menu; the actor answered it Defer and the assessment then judged that scope question as the rigor decision. The actor now answers that menu Keep and assesses the next one. * test: attribute quoted prior-record field lists, state the judge reason bound in its schema, move split-overflow to marathon Census 36629958451 reds: - outside-plan-disabled-no-fallback: the model quoted the pre-existing record as a parenthesized field list with its exact timestamp; attribution now requires that exact timestamp and the record's own field values. - plan-devex-peer-comparison-classification: the judge correctly returned missing but wrote a 1069-character reason, voiding the judgment; structured outputs cannot enforce maxLength, so the bound is stated on the field. - plan-ceo-split-overflow ran 504-1188 s as one PTY flow and set the periodic lane's wall clock; it now runs weekly in the marathon lane. * test: supply holdDeferKeepIndex to the CEO routing mocks and follow split-overflow into the marathon lane The registered-callback fixtures mock ceo-mode-option and lacked the new export; the split fixtures asserted the periodic tier; the registered-budget check looked for split-overflow only in the periodic manifest. * fix(qa): checkpoint receipts print the report link for their exploration file qa-functional-webhook-report failed in two of three censuses because the report linked .qa-evidence/NNN capture folders as "checkpoints" and never linked exploration-NNN.json. The checkpoint receipt now prints link: [checkpoint NNN](exploration-NNN.json), and the functional report template says capture folders are not checkpoints. * docs: final census numbers in the v1.91.9.0 entry; file the paid-eval follow-ups * ci(evals): name the PR-comment loop's unused fields so shellcheck passes (SC2034) * fix(plan-ceo-review): tighten expansion pacing wording to fit the skeleton cap after the main merge The merged skeleton measured 80,166 bytes against its unchanged 80,150 cap. Same instructions: ask separately for each addition, in turn, with no pacing menu; lead each proposal with the felt experience, then shape, effort and impact. * fix(eval-pass-rates): match trial-outcome files by basename so Windows backslash paths are read * fix(evals): repair proof-run reds in design-consultation, document-release, design and QA fixtures - design-consultation Phase 1 asks one brief that confirms context and decides research; the confirm-only first question scored substance 2. - document-release defines ship-owned inputs, exact steps and the JSON result, and drops stale spawned-from-/ship text (judge actionability 3.67 -> 4/4/4). - plan-design-with-ui accepts the Step 0D focus menu the same way the shared picker does ("focus on specific ones?"). - plan-design-review plan-mode saves in three Edits instead of one final Write. - QA functional annotations ask for the full 40-character revision. - Outside-disabled attribution judges quoted prior-record data by its exact timestamp or a dated, pre-existing-record sentence; four captured phrasings replay clean and current claims still fail. - --case can select autoplan-dual-voice by its literal test name. * test(design): revert the three-Edit plan-mode flow A focused paid run still timed out at 300 s: the first three passes alone took 150 s of thinking. The case stays a named timeout red rather than cutting review depth. * test: accept 'review mode = X' auto-decide declarations and parenthetical scope exclusions in the shared-libs actor auto-decide-preserved: the product auto-decided HOLD SCOPE and said "Decision: review mode = HOLD SCOPE"; the grammar knew only "is" and ":". shared-libs-plan-callers: the recommended option said "(no hardening)" and the actor read "hardening" as an expansion. Both replay the captured text, keep negative controls, and passed focused paid runs. * fix(review): pass Review Army checklists by path, run research alongside dispatch, always probe the design detector; state review-log invocation and statuses in the caller fixture - review-army-perf-n-plus-one: the parent copied full checklists into agent prompts and ran web research before dispatch (290 s on a 12-line diff); 212 s now. - review-design-lite: 5 of 6 captured trials reported the detector absent without probing; the probe is mandatory and its first line is reported, and the contract credits only fake-engine rule ids the checklist never names. - review-exploratory-small-cli: the fixture never gave review-log's direct invocation or status vocabulary; the model ran it through bun and wrote status "blocked". The prompt states both and the validator rejects out-of-vocabulary review statuses. Each case passed a focused paid run after repair. * docs(changelog): proof-run product fixes * fix(ship): always run the design-lite detector probe; test(shared-libs): credit a failed first file view and deferred-reuse Skip wording - /ship design-lite: the probe is mandatory and any non-ready first line is stated, matching /review (5 of 6 captured /review trials had skipped it). - shared-libs-pr-coverage: the first PR 42 page-1 read printed only a jq error, so the one refetch is a legitimate recovery, charged to the same budget. - shared-libs-review-prior-coverage: the Skip option said a future review can "reuse it once snapshot coverage holds"; a conditional tail on the recorded decision is not product work. Captured-text regressions and negative controls. * fix(ship,qa,document-release): repair proof-run regressions and fixture gaps - ship-docsync-completion: yesterday's audit-scope result dropped the section's status, so /ship spliced one in; the section now opens with **Status:**. - ship-docsync-missing-asset: a missing section or old Ship-owned mode blocks before launch. - ship-docsync-late-result: the invocation record says prepare already saves the candidate selection (no extra Read; budget unchanged). - qa exploratory: await the method Reads before the first probe. - qa-callers fixture: quote the real review-log record template; allow the git log command plan-completion prescribes. - qa functional observer: a receipt caught mid-link(2) is checked at stop instead of failing with ENOENT (reproduced from CI). Each repaired case passed a focused paid run. * ci(image): pin Claude Code 2.1.284, the version users run Request-body capture shows both 2.1.251 and 2.1.284 send effort "high" to claude-fable-5-1; 2.1.284 adds the model's own profile. The slower 2.1.284 census was mostly API latency: its SDK-only judges were 25% slower too. Nine previously slow cases pass on 2.1.284 within unchanged budgets. * test: one owner per case id, a structural devex 0B setup rule, and correct design/gbrain actors - plan-design-review-plan-mode was registered by two files; the PTY smoke is now plan-design-review-plan-mode-smoke, and a registry test requires one owner per case in case-sharded files. - plan-devex-finding-floor: the template's 0B narrative-confirmation question is classified as setup structurally instead of timing out a Haiku assessor. - setup-gbrain-remote: the actor accepted 'skip' on the MCP-registration question the test asserts; it now accepts that question and declines others. - design-review-plugin-handoff: the fake engine cited a file absent from the fixture repo and index.html linked a missing styles.css. Captured-question regressions with negative controls; each case passed a focused paid run. * test: PTY harness handles clipped reviews and bundled setup tabs; AUQ judge uses structured output; design-consultation carve declines optional outside voices - ceo mode routing: a Submit review taller than the viewport, a setup tab bundled after the mode tab, and a clip through the mode question each hung or misread the run; the native answer is still verified after Submit. - judgeRecommendation requests a 1-5 enum schema; a malformed Haiku reply had scored substance 0 for a 4/5 brief. Judge failures now propagate. - carve section-loading for design-consultation declines the optional outside voices (a supported path) and treats DESIGN.md as the report; timeout unchanged. The Step 0E handoff defect is not fixed (0/15 samples across four wordings, none shipped) and is filed in TODOS. * test: fold the design-consultation completion replay into carve-section-sharding (test-of-test ratchet) * docs(todos): record the pre-push hook shard-order hang * test(qa-callers): disable git auto maintenance in the fixture repo (same guard as shared-libs; from #3002) * test(office-hours-attempt): the fake judge SDK response carries stop_reason like the real API (structured judge requires end_turn) * fix(qa): the caller STOP line says to await the method Reads before any probe ship-exploratory-plan-checks: the model read exploratory.md and sent a capture in the same response, before seeing the section's own await rule. * fix(qa): number the qa value-bar questions from 1 and say reproduced bugs already answer the first two * fix(qa): define evidence.json where it is built, point the preparation gate at the next section, name measured command durations in the report template Recurring qa/qa-only workflow-judge complaints in CI (clarity/actionability 3.33). * fix(plan-eng-review,review): a disallowed question tool is not headless; report kept tests only when some were skipped * fix(plan-eng-review): keep the headless-rule contract phrases adjacent * fix(evals): cut path variance at its measured sources - gstack-qa-evidence capture prints startedAt/completedAt/durationMs and, for --deadline captures, remainingMs; the functional report takes durations from them. The section clock notice asks for one clock read up front instead of one after every checkpoint (QA runs spent 7-14% of tool calls on date -u). - ship plan-completion: skip the audit dispatch when discovery already found no plan (the dispatch-vs-skip conflict produced an optional 60-100 s subagent). - materialize/checkpoint validation errors state the expected schema, so a rejected annotations file is fixable in one call instead of blocking the phase. - session-runner counts turns from the transcript when a run times out, so timeouts stop reporting 'turn 0'. * fix(evals): count timeout turns only from object transcript events * test(qa-callers): deterministic child transport, completion-time handoff reads, compact phase report The exploratory caller cases exist to prove the caller starts and bounds exploratory QA. Their native adversarial reviewer (review) and plan audit (ship plan-checks) now come from recorded child outputs instead of a live subagent, handoff freshness reads are required before completion records rather than every bookkeeping log, and the phase report is compact. Measured: 194-257 s per case against 208-284 s before, no subagent calls. * test(ship-docsync): seed fault cases at their gate instead of replaying attempt 1 The post-dispatch fault cases (missing-marker, launch-failure, timeout-unsettled, late-result, stale-before, stale-after, recovery) now start from a fixture-owned attempt 1: the real actor prepares and dispatches it, its verbatim output is saved once, and the invocation journal carries its pre-dispatch entry with the child asset hashes. The model resumes at Parent processing with a trimmed read list, inspect named as the authoritative repository observation, and recovery's intermediate checkpoint folded into the next attempt's pre-dispatch entry. Assertions count only parent-issued transport events and require a read of the saved attempt-1 output; missing-asset and the legacy failure case keep the full model-driven first attempt, and their prompts are byte-identical. * test(ship-docsync): name the seeded read list and cap journal/report length The first seeded stale-before run spent calls locating documentation.md (two ls sweeps), reading through cat and re-Reading the record before Edit, and ~40 s composing 1.5-2.2 KB entries and report. Name every seeded read path, ask for native Read, and bound entry/report length. * test(ship-docsync): trim the seeded parent's measured model time Measured on the seeded runs: one read the 78 KB ship/SKILL.md, the post-child freshness comparison spent 18-32 s of thinking over full inspect contents, and the final response restated the report (~1.1 KB). Say the phase excerpt stands in for ship/SKILL.md, compare hashes first and read content only for changed paths, and end with one status line. * feat(qa-evidence): enforce the checkpoint sequence and fill report bookkeeping in code - capture refuses to run another probe until a checkpoint anchored on the latest complete capture names this capture as its next command, and every complete capture prints that requirement. - materialize fills revision, runtime, cwd and learning (checkpoints whose next native command differs) when omitted and prints the reportLinks the report must include; the QA section shrinks accordingly. * test(qa-callers): hand the caller phase its invocation-start observations and review token; fix(next-version): fetch without auto maintenance - Every caller case receives the diff, status, log, untracked list, HEAD and an already-captured review start token, so the phase spends its budget on the contract under test instead of re-running setup reads. - gstack-next-version's fetches pass --no-auto-maintenance. On git 2.55 a completed fetch forks detached maintenance in the caller's repository; the free suite's live smoke test ran it inside the CI checkout, and every shard-12 pre-push hook hang so far followed a completed smoke fetch. * feat(deslop-shared-libs): route every Git read through bin/gstack-safe-git The skill made the model retype a long safe-Git prefix on each call and a dropped flag failed shared-libs-read-only. bin/gstack-safe-git applies the fixed env + flag prefix, adds --no-ext-diff --no-textconv to log/show/diff, allows diff only between two explicit object IDs and ls-files only in the NUL-delimited overlay form, and refuses every other shape with one line naming the allowed forms. The template now points at the installed helper (host global runtime via {{SAFE_GIT}}) and drops the prose it enforces. Fixtures resolve the helper to this checkout, the git shim records the safety environment, and isGuardedGitRequest requires the complete prefix (env included) for every repository read. * test(shared-libs): tee to a discard device is not a file write Paid shared-libs-opportunity-judgment t1 on 1213b01 failed read-only on '... | tee /dev/null | sha256sum'. The detector flagged any tee operand while the same devices are allowed for redirection. tee now fails only when an operand is a real file; tee to a file, -a file and -- -a stay violations. * fix(qa-evidence,observer): reject placeholder metadata and replay-only learning; declare the docs atomic-write target - materialize measures revision, runtime and cwd itself and rejects supplied values that differ (CI run wrote revision "HEAD" and runtime "bun"), and refuses learning checkpoints that replay the same probe, naming the fix. - The docs write observer treats Claude Code's atomic temp for the authorized doc target as transient, so a temp renamed before its per-file watch no longer marks the observation incomplete (ship-docsync-completion flake). Per-file monitoring outside declared targets stays fail-closed. * test(qa-functional): fix mode requires only the happy scenario from the model (carried byte-identical from #3002 183b01f4..3e6074b4) verifyQANativeRegression already reruns all eight webhook scenarios on the repaired source, so the model-side eight-scenario requirement in fix mode duplicated harness coverage and pushed qa-functional-webhook-fix past its budget. qa-only still requires every scenario. * fix(deslop-shared-libs): probe the audited repository with -C <repo> A CI run probed safe-git from the session directory above the target repo, so the capability probe never touched the repository and the run fell back to the API without a local attempt. The probe (and any call from elsewhere) now names the audited repository. * test(qa-deadline): never attach a reader to the full-pipe fixture's stdout The full-pipe receipt test attached a 'data' listener (flowing mode) and then paused; on CI the reader could drain the 2 MB write before the pause, so the receipt write never blocked and the helper exited 0 in ~126 ms. The stdout pipe now stays unread until the assertion, which is what the test means to model. * feat(qa): helpers answer --help, and the QA eval interfaces declare it Approved by Garry: asking gstack-qa-evidence or gstack-qa-deadline for usage is read-only, so both helpers print usage and exit 0 on --help (the evidence usage now names the annotation shape), and the functional and caller command allowlists accept exactly 'bun <path>/bin/gstack-qa-{evidence,deadline} --help'. Two CI runs failed only on that call. * fix(qa): after an input change, a probe is affected unless shown otherwise CI late-input run finished in time but revalidated only the happy probe after the locale input changed and reported the stale adverse probe green. The revalidation step now treats any probe not shown to be unaffected as affected. * test(shared-libs): seed the lifecycle replay's first Step 3 pass instead of replaying it shared-libs-review-lifecycle ran ~88% of its 300 s session budget (12-run census median 265 s, 4/24 sessions timed out). The fixture now executes pass 1's Step 3 once with the real logger and Git: a real unused REVIEW_START, then the diff, inventories, attributes/config/index flags, gstack-review-read output and every file's bytes and sha256, saved to one observation. The model resumes at Step 4 with an exact four-file first read, the observation named as the authoritative pass-1 repository read, one post-fix verification, an explicit pass-2 read list and a twelve-line summary. Pass 2 still runs its own --start, diff, reads, fingerprint and stage actor before --finish. The actor scope now states that a current settled final-pass actor result supplies the replaced QA/adversarial prerequisites and that the no-credit disclosure is a reporting label: one r1 session persisted completed:false from that ambiguity. New assertions: the final binding never uses the seeded token's start or tree, and the observation was read; free controls finish the seeded token (binding changed) and omit the observation read, and both fail. * test(shared-libs): trim the resumed review replays' setup and report Every sibling review session (revalidation, path-eligibility, index-flags, prior-coverage) loaded qa/sections/exploratory.md and often scope.md although its QA and native adversarial results are supplied synthetic inputs, then spent a second request on shared-code-reuse.md and base metadata. The resumed scope now states that the supplied results replace Step 4's QA method loading; the revalidation contract names one first response (workflow, checklist, finding, prerequisites, shared-code-reuse.md, base metadata) and caps the summary at twelve lines. Receipt order, direct source reads, the checker, the question and final persistence are unchanged. * fix(review): define what a Step 5c Skip option says Step 5c named "B) Skip" without saying what its description may claim. Two CI captures (path-eligibility on 131d43be, index-flags on 4643cb85) offered a Skip whose description added effects beyond declining: "The extraction can be applied in a later editing review pass" and "replacing the invalidated prior Skip". Those read as change commitments, so the no-change actor refused both. Step 5c now says to describe Skip only as no code/index change with the Skip recorded; adjacent lines are compacted so the review parity caps hold unchanged. Both exact packets are kept as a free regression: still refused, and accepted once Skip follows the rule. The actor's classifier is unchanged. * fix(qa-evidence): every complete capture needs an evidence row; test(tpa): accept the hyphenated app-specific-password spelling - materialize refuses when a complete capture has no evidence row and is not named in limits (CI cli-report omitted capture 004), naming the missing IDs. - tpa-apple-ban's detector required 'app-specific password' with a space; the CI answer said 'app-specific-password path' and was otherwise correct. * test(qa-observer): fix mode treats atomic temps of authorized src/test writes as transient CI webhook-fix failed with 'Could not watch test/worker.regression-1.test.ts.tmp...': Claude Code's Write renamed its temp before the per-file watch was added. The functional eval now tells the observer its mode, and a temp whose target that mode may write is observed through its directory watch. Report-only mode and undeclared paths keep failing closed. * feat(qa-evidence): refuse evidence observed on an older input snapshot than the latest capture When native probe output declares a top-level input snapshot, materialize compares each evidence row with the latest capture's snapshot and refuses stale rows unless they are classified superseded, naming the captures to rerun. ship-exploratory-late-input kept reporting a pre-change adverse probe green after the input changed. * test(qa-functional): point the fixture at the helper's --help instead of its source A CI webhook-fix run spent three turns reading lib/qa-evidence.ts to learn the interface and timed out just before materialize (agreed with #3002's owner). * feat(qa-evidence): captures list the caller's declared-but-unrun required probes GSTACK_QA_REQUIRED_PROBES (a JSON array of native child commands) makes every capture print requiredRemaining; it never judges pass or fail. The functional eval passes the webhook list from QA_WEBHOOK_REQUIRED_SCENARIOS, which the verdict now reads too, so the nudge and the verdict share one source (agreed with #3002's owner). CI webhook-report kept stopping with scenarios unrun. * test(review-army): record N+1's pre-dispatch stages and scope the session to Step 4.5 review-army-perf-n-plus-one timed out in 7 of 13 CI runs on this branch (passing 245-280 s of 300). Each session spent ~95 s on setup (the full extracted SKILL, checklist, section greps, exploratory.md, diff-scope/stats/learnings, tooling checks), ran Step 4's core pass, a search-before-recommending WebSearch, and wrote a 10-16 KB report (~100 s after the Red Team returned). The fixture now stages only review/sections/review-army.md plus the performance and red-team checklists, and hands the session the recorded detect-scope, specialist-stats and learnings outputs and the diff. The caller passes --performance (every CI parent already treated the prompt as that force flag against the <50-line skip), declares the core pass, QA, adversarial review, web research, Fix-First and persistence out of scope, and caps the report at the selection line, the SPECIALIST REVIEW block and the Red Team result (30 lines). The Performance specialist and the conditional Red Team are still real foreground subagents, and the report still has to surface the N+1. New assertion: a foreground Performance specialist dispatch precedes the Red Team dispatch. Free controls omit the Performance dispatch or background it, and both fail; the budget lifecycle adapter supplies the current result shape. Touchfiles now include the .rb fixture the case reads. * test(review-army): share the recorded Step 4.5 staging with consensus and supply its Red Team review-army-consensus (periodic) timed out in 2 of 13 census sessions; passing runs took 213-297 s of 300. Like N+1 it spent ~30-50 s reading the whole extracted SKILL, checklist and every specialist file, sometimes dispatched an unrequested Maintainability specialist, then ran a Red Team (60-70 s) and a second merge before writing a 9-15 KB report. The N+1 staging and scope text move into stageReviewArmySession / reviewArmyScope / reviewArmyChecklists (the N+1 prompt renders byte-identical). Consensus now records its detect-scope, stats, learnings and diff, stages the Review Army section with the security and testing checklists, forces --security --testing, and caps the report like N+1. Its Red Team is outside the multi-specialist contract, so the fixture supplies a labeled synthetic NO FINDINGS result instead of a dispatch. The existing SQL-finding and browser-error assertions are unchanged; the lifecycle adapter's spawnSync now returns the git output the staging reads. * docs(changelog): v1.91.10.0 records the flake census and its repairs * test(strict-output): give the spool-prefix child time to finish before the pending stream times out windows-free-tests failed on 9a7a7e54: the 150 ms shared deadline raced Bun startup on Windows, so the child was killed mid-write and the spool held a partial payload. Only the never-released extra stream should time out; the child now has 3 s. * fix(qa-evidence): accept a single limits string; test(qa-callers): read the handoff first when a probe snapshot changes CI late-input spent a turn rewriting limits as an array after materialize refused a string, and a ten-read sweep hunting for the changed input before it read reports/HANDOFF.md, then timed out at 300 s. * test(autoplan-dual-voice): unwrap the framed native report before Claude Code 2.1.284's agentId/usage trailer * test(section-loading): credit a Bash print that contains every line of the carved section * test(auto-decide): ask for the selected mode in the skill's mode handoff line, not a separate public decision * test(plan-ceo floor): scope preservation approves no premise, approach or remedy * test(autoplan-dual-voice): the fixture declares that delivered bash blocks run alone, diagnostics separately * test(coverage-audit): a fenced plain-word caption in a successful && read chain is display only Census 36776104571 plan-eng capture read both owned files with cat -n in one successful && chain; the caption 'echo "=== git diff main --stat ==="' fell outside the two-token caption grammar, so both reads lost credit. Accept a fenced caption of plain words; unfenced command strings, expansions, redirection, -e escapes and ; / || tails stay rejected. * test(office-hours): a fork whose outer options are the seeded shapes is the Phase 4 question Census trials 1-2 captured complete Phase 4 forks (A) Server-side B) Client-side C) Hybrid, recommendation with because) whose prose used none of the vocabulary words. Accept two seeded shapes as outer options as Phase 4 specificity; the earlier-phase, nested, fenced and single-shape controls still fail. * fix(review): design-lite rows keep the detector's [rule-id]; the e2e detector rows point at the diff The output template had no rule-id slot, so rows merged with checklist items dropped the detector id (census t2, local t1). Rows now carry [rule-id]. The fake engine's sample rows named a foreign fixture path at line 0; the e2e remaps them to landing.html/styles.css so trials stop spending turns reconciling it. * test(shared-libs): the plan actor reads scheduler parity and unchanged-scope lists Census 36776104571's question preserved the contract ('behaving exactly like the scheduler', 'scheduler parity holds by construction') and excluded work with 'Existing copies and helper hardening stay unchanged'. Accept exactly/parity as preservation (negated forms refuse) and a bare noun list that stays unchanged as an exclusion for the expansion scan only; verb-led clauses still refuse. * fix(qa-only,qa): name the exploratory read point and finalization order; judge qa with its browser assets qa-only judges cited 'next section' pointing at the wrong heading, an exploratory trigger that contradicted its read point, clock ownership in mixed runs and the unstated order of exploratory section 4 vs reporting. The qa judge penalized the absent qa-report-template and issue-taxonomy that qa-patterns loads; with them in, it found issue-taxonomy's dangling 'rule 13' (the consent rule is browser rule 3). * test(ship-docsync): seeded attempt 1 counts toward the limit; transport counts ignore calls that never reached the state file - CI launch-failure retried after the seeded attempt 1 as if that attempt were the fixture's; the seeded prompt now says attempt 1 is this invocation's and a further attempt needs what Blocked recovery requires. - A late-result run typo'd the state path once (ENOENT, the actor never ran), then repeated the call correctly; the per-action count compared both calls with one actor event. Only calls naming the real state file are counted. * fix(plan-eng-review): show the accepted dedicated read form for coverage-diagram sources CI plan-eng-coverage-audit mixed package/config and git diff into the source read; the review variant, whose prompt shows the && display form, does not. The plan trace step now shows it too, within the unchanged size cap. * test(sync-gbrain-readiness): a negation earlier in the claim clause is not a search/write readiness claim The census unknown actor wrote 'nothing about read, search, or write capability is confirmed either way' after a YELLOW/WARN verdict. The claim window started at 'write', so the leading 'nothing' was outside it. Check the clause subject for nothing/neither/none/no; keep the original in-claim negations. Replay of the captured output passes; positive controls still flag an unnegated claim. * fix(office-hours): a forcing question's recommendation takes the position the founder's words support auq-matrix office-hours asked D1 Demand as options about the founder's own evidence and, with no rule for that shape, recommended 'answer whichever is TRUE — A is marked recommended only because it is the strongest position' (substance 2). Say what such a recommendation is: the option the founder's own words support, why it matters for the next step, and what would change it. * fix(plan-ceo-review): name the mode preference command and the exact handoff line auto-decide-preserved at 6fcb0981: the model never ran the preference check, read 'check ... through the preamble' as already done, auto-selected 'per your preference setting', and wrote 'Selected mode: HOLD SCOPE, auto-decided from your tuned preference' instead of the AUTO_DECIDE handoff line. At 9a7a7e54 it ran the check but wrote 'Decision: HOLD SCOPE is the review mode for ...'. Neither matched the handoff template the observer recognizes. Name gstack-question-preference --check at the point of use and say the handoff begins with the exact matching line. Collapse the audit block's comment padding to stay within the unchanged 80150-byte skeleton cap. * test(section-loading): record the CEO capture's report and transcript The 6fcb0981 census failed hasStaleFillRaceFinding (line 98), but the case records nothing beyond junit, so the report the detector judged is gone. Return the SkillTestResult from captureSectionReads and record it, with the full saved report, through the eval collector on pass and fail. * test(design): plan-mode names its read list and caps its additions and summary At 6fcb0981 plan-design-review-plan-mode timed out at 300 s (9 turns): 22 cat/sed chunk reads (~50 s), then a 28 KB plan Write (~150 s), before the read-back finished. The 9a7a7e54 pass took 240 s with a 24.6 KB Write. Read SKILL.md, review-sections.md and plan.md natively in one response, keep additions under 14,000 characters and the summary within ten lines. Budgets unchanged. * test(plan-mode-no-op): require prose evidence before a waiting verdict ends eng/design runs (carried byte-identical from #3002) With the prose fallback forced, the gate renders as a lettered menu; a judge 'waiting' verdict on a spinner-only frame ended the run as 'asked' before the menu rendered, so the scope-gate check failed on unchanged behavior. * feat(qa-evidence): materialize computes the phase verdict; callers must report it Approved by Garry: the helper, not the model, decides whether evidence can pass. materialize writes verdict {status, open} into evidence.json and prints it: fail or blocked from row classifications, inconclusive while any row is superseded, a complete capture is withheld, a declared required probe is unrun or there is no evidence, else pass. The caller fixture requires receipt.status to equal that verdict. CI late-input kept reporting pass with a superseded happy probe. * test(qa-callers): compare the receipt with the helper verdict only when evidence.json was materialized The producer free tests run captures without materialize; evidence.json is optional for callers, so its absence is not a verdict mismatch. * test(llm-judge): run the ship workflow judge at medium effort so its panel fits JUDGE_MS claude-fable-5-1 accepts only adaptive thinking (thinking.type.enabled with budget_tokens returns 400), so effort is the available thinking control. Measured on the exact ship judge request (105,301 input tokens): - default effort, 18 samples: thinking 5,086-10,881 tokens, 75.9-144.7 s; 3 of 18 passed the 120 s deadline (about 42% of 3-sample panels). - medium effort, 18 samples: thinking 2,749-5,762, output at most 6,144 tokens, 43.1-77.9 s; scores 4/4/4 in 16 of 18 (clarity 3 in two), versus 14 of 18 at default. callJudge gains an effort option sent as output_config.effort; only the ship judge sets it. Rubric, floors, panel size, deadline, model and max_tokens are unchanged. The cache identity records effort. * test(llm-judge): ask frontier workflow judges for 120-word reasoning under the unchanged 150-word check Told "under 150 words", the ship judge's reasoning landed at 130-156 words (3 of 18 probe samples at 152-156), so the structured-response check failed about one panel in three independent of effort. The prompt's frontier block and the response schema description now say under 120 words; the validator still rejects 150 words or more. The changed prompt bytes reach only the two frontier judges: ship/SKILL.md workflow (prompt and schema) and review/SKILL.md workflow (prompt). * test(llm-judge): type the stream transport mock call * test(plan-ceo floor): the request answers only the questions it names PR lane 36794871032 (head 20d6e98f): the CEO floor ran 608 s without a question. Its Step 0 recorded the premise gap and approach choice as unresolved ledger rows, then said "this session supplies all answers up front, so no decision brief was dispatched" and wrote Sections 1-11. 2734e203 stopped scope preservation from approving the premise; this time the actor block (declined setup, recall, outside voices, HOLD SCOPE) and the fixture's "complete user request is available from the start" were read as pre-answering every review question. The CEO actor now states that the request answers only the routing, recall, outside-reviewer and review-mode questions it names. * test(plan-devex floor): a 'Partly wrong' 0B answer is the narrative confirmation PR lane 36794871032: the DX floor asked its D1 narrative confirmation (Accurate, proceed / Partly wrong, correct it / Way off, actual is...). The deterministic setup rule accepted only 'Some ... wrong', so the question went to the LLM assessor, which hit its 30 s spawnSync cap (ETIMEDOUT) and ended the case as assessment_error at 141 s, the same failure as census 36641820398. The rule now accepts 'partly' beside 'some'; the captured question is a free regression and the remedy-option controls still go to the assessor. * test(design-review plugin handoff): quoted report text is not an install command PR lane 36794871032: every behavioral check passed except noInstallOrOverride, which matched "no `npx impeccable`" inside the quoted heredoc that wrote detector-output.md. Nothing was installed or downloaded. The check now drops quoted-delimiter heredoc bodies (literal data) before matching; unquoted bodies, which can expand $(...), and unterminated bodies stay checked. Free controls cover the captured write, bare npx, an IMPECCABLE_BIN override, an unquoted $(npx ...), npx after the delimiter and an unterminated body. * test(review-army delivery audit): stage only the plan-completion section and record its git reads PR lane 36794871032: the case timed out at its 120 s budget after 7 turns (previous lane passed in 45 s). The session read the 46 KB extracted SKILL in three passes (cat to persisted output, grep, sed), ran its own git reads, wrote a 74-line report, then inspected and ran gstack-learnings-log and rewrote the report's Learnings section. As in the Step 4.5 cases (17ee2e54/2bd4651c), the fixture now stages only review/sections/plan-completion.md, hands the session the recorded git log and diff, declares the HIGH-impact question, its Scope Check, learnings logging and later steps outside the capture, and caps the report at the audit block and its DISCREPANCY entries (30 lines). The NOT DONE and email assertions are unchanged. * feat(qa-evidence): one capture call records the causal note for the previous capture capture R NNN [--public] (--deadline D|--timeout-ms MS) --after PREV --hypothesis 'TEXT' -- CMD publishes exploration-NNN.json {observationCapture, observationArgv, observed, hypothesis, nextCapture, nextArgv} before running CMD, refusing unless PREV is the latest complete capture. The receipt carries checkpoint/checkpointSha256; validators bind the note to the transcript's capture calls by capture ID and receipt hash instead of exact command strings. The separate checkpoint command and the capture guard keep working; materialize learning accepts both note shapes and still rejects same-probe replays. Prose and eval fixture prompts teach the merged form. * fix(qa-evidence): a superseded row stops holding the verdict open once its probe is rerun on current inputs materialize requires an old-snapshot row to be classified superseded, and its verdict kept every superseded row open, so rerunning the probe (what its own error tells the model to do) could never reach pass; late-input reran 3 and 9 on the new snapshot and still got inconclusive. A superseded row now closes only when a non-superseded row with the same captured argv observed the current snapshot. Re-materializing an already-published evidence.json names the cause instead of failing generically. * test(plan-eng batching): count saved decisions whose label drops the (recommended) marker or whose report is titled 'Eng Review Report — <plan>' * fix(qa): browser-only runs skip annotations/materialize; only Q captures can anchor evidence rows * test(design): plan-mode length is a drafting target, not a check to measure and trim * test(llm-judge): structured output for doc, outcome and posture judges so reasoning quotes cannot break JSON * test(ship-docsync): steer skill file reads to Read; large cat output becomes an unpageable preview * docs(changelog): browser-only QA evidence and structured judge output * test(qa-only cleanup): refusal scenarios get a 1 s budget and an absolute worker deadline; 300 ms starved under parallel load * fix(office-hours, design-consultation): ask the goal question and read the mode section first; ask the memorable-thing question on its own * test(outside-disabled): a record named by the retained record's own clock and then disowned owns its completed status * test(context-skills): install gstack-paths in the fixture bin; without it the model guessed the checkpoint root * test(ceo mode routing): SCOPE EXPANSION posture credits plural 'expansions' * test(ship-docsync): name the unmet atomic-replacement check on a forbidden temp-file write * fix(qa): browser-only runs materialize an empty evidence list with checkpoints in limits, matching /qa-only * test(qa callers): an accepted review-log record may cite checkpoints as finding evidence * fix(plan-eng-review): state that a disallowed question tool never qualifies as headless before the headless action * merge follow-up: re-record paid CLI parity for #2999's flags; trim merged review, qa-only and plan-eng wording toward the size caps * test(golden): refresh codex/factory ship goldens for the trimmed caller QA wording * test(coverage-audit fixture): disable git auto maintenance so cleanup is not racing a detached git writer * test(parity): raise review, qa and plan-eng caps to the measured merged size of #2999 and #3002 (each fit alone), documented per cap * fix(qa-evidence): materialize rejects an unrecognized classification before publishing, so the one-shot verdict cannot be locked inconclusive by a descriptive label	2026-10-01 13:55:16 -07:00
Garry TanandOpenAI Codex	4a3c6a8a3c	v1.87.0.0 feat: add verified CSO audits and replayable repair bundles (#2852) ... * feat(cso): add verified audits and replayable repair bundles * fix(cso): harden qualification and setup boundaries * fix(cso): assemble security canaries at runtime * fix(cso): bound release proof and maintenance work Co-Authored-By: OpenAI Codex <noreply@openai.com> * fix(cso): require complete evaluation reports Co-Authored-By: OpenAI Codex <noreply@openai.com> * fix(cso): replay expired snapshots from supplied source Co-Authored-By: OpenAI Codex <noreply@openai.com> * test(cso): synchronize DNS cancellation assertion Co-Authored-By: OpenAI Codex <noreply@openai.com> * chore(ship): exempt repository owner from liveness proof Co-Authored-By: OpenAI Codex <noreply@openai.com> * test(cso): make recheck retention overlap deterministic Co-Authored-By: OpenAI Codex <noreply@openai.com> * chore: bump version and changelog (v1.85.0.0) Co-Authored-By: OpenAI Codex <noreply@openai.com> * fix(cso): pass native release gates Co-Authored-By: OpenAI Codex <noreply@openai.com> * chore: move release to v1.86.0.0 Co-Authored-By: OpenAI Codex <noreply@openai.com> * fix(cso): resolve rechecks by finding Co-Authored-By: OpenAI Codex <noreply@openai.com> * chore: move release to v1.87.0.0 Co-Authored-By: OpenAI Codex <noreply@openai.com> * fix(cso): pass macOS and Windows release gates Normalize BSD wc output, compare Windows paths by filesystem identity, preserve portable snapshot race coverage, and narrow POSIX-only Windows fixtures. Co-Authored-By: OpenAI Codex <noreply@openai.com> * fix(cso): harden native verification gates * fix(cso): refine Windows native diagnostics * test(cso): isolate Windows Git startup failure * test(cso): stabilize Windows native diagnostics * fix(cso): support hardened Git on Windows * fix(cso): close final verification gaps * test(cso): bound cold Docker fixture setup * fix(cso): restore cross-platform free-suite gates --------- Co-authored-by: OpenAI Codex <noreply@openai.com>	2026-09-14 15:14:58 -07:00