3 Commits

Include renames

Author	SHA1	Message	Date
Garry Tanandcapy-ai[bot]	4c6fa89cdc	v1.91.159.0 security: gstack's Bun entry points ignore the audited repo's .env and bunfig.toml (S1) (#3312) ... * security: isolate every gstack Bun entry point from the audited checkout's .env and bunfig.toml Bun auto-loads .env, .env.local and .env.<NODE_ENV> from its cwd and runs a cwd bunfig.toml preload before the program starts. gstack runs inside the repo it audits, so a committed .env could steer state roots, the claude binary or spend caps, and a bunfig preload ran arbitrary code in every bin and hook. Every bun shebang, shell and skill caller, TypeScript spawn and compiled build now disables both (--no-env-file --config=/dev/null, bunIsolationArgs(), --no-compile-autoload-dotenv/bunfig), with a ratchet and hostile-fixture regressions. Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * test: name the directly executed scripts in the isolation ratchet Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * test: follow the isolation flags in touchfiles, copy lists, recipe assertions and the harness pin Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * fix: keep browse/src/bun-child-config.ts unchanged so the native cookie qualification still binds Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * fix(windows): the empty bunfig outlives its process so the timeline hook's detached repair child can read it Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * v1.91.159.0 security: gstack's Bun entry points ignore the audited repo's .env and bunfig.toml (S1) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * test: plan-mode render pattern accepts the isolation flags Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> --------- Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>	2026-10-20 00:24:28 +00:00
+4454 people	2db0b3adc8	v1.91.19.0 fix: severe fix wave (truthful gates, no lost memory, every host finds gstack) (#3025) ... * docs: approved severe fix wave plan (Oct 2026) * D0: keep a project's .env and bunfig.toml out of compiled binaries and daemon children Compiled binaries are built with --no-compile-autoload-dotenv and --no-compile-autoload-bunfig; the browse server and terminal agent are spawned with --no-env-file and an empty config, so a repo's .env cannot set GSTACK_CHROMIUM_NO_SANDBOX and a bunfig preload cannot run code in them. Verified on Bun 1.4.0. * A9: --no-write is a dry run and leaves ingest state untouched --no-write stamped every prepared page ingested_at and saved state, so a later real run skipped those pages forever. It now prepares and reports pages without importing or changing the state file. Tests that used GSTACK_MEMORY_INGEST_NO_WRITE / --no-write to seed state move to an explicit seam (a fake gbrain that accepts every staged page), because the --no-write contract deliberately changes. * H #2885 #2827: vector geometry is not phone or card PII pii.phone.e164 accepts "." as a group separator, so a single float (viewBox="0 0 581.66796875 ...") and a run of path coordinates ("37.6188 101.694") read as phone numbers. pii.cc's \b stops at a decimal point, so the long fraction of an .excalidraw float is a Luhn candidate. Excalidraw's seed/versionNonce/updated integers read as bare phone numbers. A rendered /diagram triplet raised 18 (svg) and 97 (excalidraw) MEDIUM findings on main; it now raises none. - looksLikeDecimalCoordinates (ported from #2885, extended to a single decimal and refusing a leading "+", the E.164 marker) - insideDecimalNumber: a cc span glued to "<digit>." or ".<digit>" - isNumericMetadataValue: an unquoted JSON value under a seed, nonce or timestamp key; the same digits under "phone" still report Controls: E.164 in prose, dotted and spaced phones, "+"-prefixed runs, "phone"/"contact" JSON numbers, Luhn-valid cards in prose and at sentence end. Co-authored-by: Lewis <lewispeel@hotmail.com> * D1: pin the terminal-agent /ws Origin to the gstack extension (#1324) GSTACK_EXTENSION_ID moves to browse/src/extension-id.ts, shared by /extension-token and the terminal agent's /ws gate. /ws now returns 403 for every origin except the pinned extension. Forks override the ID with gstack-config set browse_extension_id <id>, honored by both routes; the BROWSE_EXTENSION_ID env var is no longer read. Test sites that used fake extension IDs now use the pinned origin. * H #2899: env.kv skips a function call assigned to a credential name `session = _FlakySession(...)`, `token = make_token(user, ...)` and `password = getpass.getpass()` raised env.kv MEDIUM, which blocks the fail-closed decision log. The exemption needs the line's evidence, not the span's shape: the value is unquoted, is an identifier or dotted path followed by "(", and the call closes on the line or opens a multi-line argument list. Quoted values, unclosed "(" inside an unquoted secret (API_SECRET=Abc123xyz(9Qm) and calls carrying a high-entropy literal argument still report. The literal check is shared with the #2912 env-read exemption. Co-authored-by: Yugz <yannduzelier@gmail.com> * H #2962: per-machine config filenames are not internal hostnames `CLAUDE.local.md` and `settings.local.json` reported internal.hostname MEDIUM (the pattern's \b stops at the dot before the extension), which fails closed in the decision log. A span immediately followed by a known config extension (.md .json .jsonc .yaml .yml .toml .ini .conf .txt) and a word boundary is a filename. Controls: `ssh build.local`, `printer.local.` at a sentence end, `printer.local/md`, `printer.local.mdx`, a host with a port, `api.corp`, `build-7.internal`. The `.env.local` vectors (#2740) pass on main already via isDotenvFilename and are pinned here as regressions. Co-authored-by: Florian <florian.cleyetmerle@gmail.com> * H #2784: four-part versions need a version declaration, not a shape `"version": "1.128.1.0"`, `<Version>`, `AssemblyVersion("1.0.0.0")` and Keep a Changelog headings raised pii.ip_public MEDIUM on every release, because a four-part version is byte-for-byte a dotted quad. Exempt only when the same line declares it a version immediately before the value (a version key or word, an XML tag, an attribute call, or a `## [x]` heading). The same digits after host:/server =/"upstream", bare on a line, or after prose that merely mentions a version still report. Unlike #2784, this does not exempt a value because the same string is declared a version elsewhere in the text: that let a real address sharing a diff with a matching version slip. The pre-push hook's VERSION-file rule (#2856) is unchanged. Co-authored-by: cdecook23 <decook23@gmail.com> * H #2913: postgres:postgres is exempt only on a loopback host The postgres image's default pair in compose files and CI DATABASE_URLs blocked pushes as db.url_with_password HIGH. It is now exempt only for postgres/postgresql URLs with exactly postgres:postgres on localhost or 127.0.0.1 (optional port). A bare single-label host (`db`, `postgres`), any other host, a host that merely starts with localhost (`localhost.evil...`, `localhost@prod-db`), any other user or password, and other schemes still block. #2913 exempted the pair on any host; this keeps the plan's narrower rule. Co-authored-by: mb1810 <michael.berger1810@gmail.com> * G1: ship without a version when no version source is configured Resolve the version source to four outcomes (valid, absent, ambiguous, broken) in lib/version-source.ts and read that one signal in gstack-version-bump (new NO_VERSION state; write refuses to invent a VERSION), gstack-next-version (no_version, no slot), /ship's bump, CHANGELOG and title steps, /land-and-deploy's drift check and /document-release. A configured version file that is missing, empty, unreadable or malformed stops with its path; 0.0.0.0 is never substituted. docs/skills.md documents .gstack/version-path. Migrates two tests that pinned the old fabricated-0.0.0.0 behavior. Co-authored-by: Koda <koda@edgeviewfinance.com.au> Co-authored-by: Merit Blake <merit@newsprint.com> * H #2949: eval credential screen checks every name segment The hermetic eval env admits GITHUB_* and EVALS_* metadata but rejects credential-shaped names, judged only by the last underscore segment. A trailing qualifier carried a credential through: GITHUB_APP_PRIVATE_KEY_BASE64, GITHUB_TOKEN_1, EVALS_API_KEY_FALLBACK reached eval children. The screen now rejects a name when any segment is a credential word; segments, not substrings, so GITHUB_PATH / GITHUB_TOKENIZER / GITHUB_KEYRING stay metadata, and explicit runner admissions still win. Co-authored-by: Nguyen Thanh Dat <ntdat812@gmail.com> * H #2915: cover chunked scanning of large multi-file push diffs Test-only coverage for behavior already on main: a clean two-file pushed diff over the engine's 1 MiB cap passes without engine.input_too_large, and a credential at the end of the second file (a later scan chunk) still blocks the push. Co-authored-by: pneumorea-ai <pneumorea@gmail.com> * A3: keep a project's DATABASE_URL away from a PGLite brain With a PGLite config, buildGbrainEnv returned the caller env unchanged, so a project's .env DATABASE_URL (autoloaded by Bun) reached gbrain: a false broken-db at best, gbrain talking to the app's own Postgres at worst. Inherited DATABASE_URL and GBRAIN_DATABASE_URL are now blanked for PGLite configs (GSTACK_RESPECT_ENV_DATABASE_URL=1 still keeps them), and every gbrain child runs from gbrain's config dir instead of the user's project, so it cannot reload the project's .env / .env.local. The old test that pinned pass-through for an engine:pglite config is migrated: pass-through now applies only to configs that name neither a database_url nor PGLite. Co-authored-by: maxpetrusenkoagent <max.petrusenko.agent@gmail.com> * C9: guard destructive gstack-upgrade and spec fences against unset or wrong paths Fresh-shell hosts (and Claude Code, which keeps cwd but not shell variables) ran cd "$INSTALL_DIR" with an empty variable, so git checkout/stash/reset targeted the user's project. Step 2 now prints an absolute path; every destructive fence starts with cd -- "${VAR:?...}" || exit 1 and checks it is gstack's own checkout before any git command. /spec's worktree and spawn fences get the same guard. * F4: keep every paid design variant across regenerate rounds $D variants now reserves a collision-safe round-<N>/ directory per run (atomic mkdir) and writes a manifest.json of what each round saved. The paid image request and the local write retry separately: a write failure retries the write (temp + rename) and, if the path stays unwritable, keeps the bytes in the temp dir instead of buying the image again. Responsive variants land in their round directory too. Readers updated: design-shotgun reserves a round dir and a private temp dir for its parallel agents and records approved_path/round in approved.json; design-html resolves approved_path (legacy letters still work) and finds round-dir variants; design-consultation checks the printed paths; the gallery shows round-dir variants. Variants that were already overwritten before this change cannot be recovered. Migrates the consultation contract test that pinned the flat path. Co-authored-by: Matt Van Horn <mvanhorn@gmail.com> * A2: stage dot-leading slugs gbrain can collect; network errors are db-unreachable A page slug segment starting with a dot (a .claude project) staged a file gbrain's import walker prunes, so the batch was refused on every run. Leading dots become "dot-" per segment, and slugs recorded in state under the old name map through the same function so no page is duplicated. A DNS or connection failure (ENOTFOUND, EAI_AGAIN, ECONNREFUSED, ...) no longer classifies as broken-db, which routed users to move a working config aside. The new db-unreachable status carries the reason "database host unreachable (<code> <host>); your gbrain config is unchanged. Fix: check network or VPN, then re-run /sync-gbrain.", skips local sync stages with that reason, and counts as usable for gstack-gbrain-detect --is-ok like timeout. Co-authored-by: Infiniteyieldai <infiniteyield.ai@gmail.com> * D2: one address classifier for navigation literals, DNS answers, redirects and page-driven navigations (#2811) classifyAddress() in url-validation.ts blocks 169.254.0.0/16 (metadata and container credential endpoints), 100.100.100.200, fe80::/10 and fc00::/7, including IPv4-mapped, IPv4-compatible and NAT64 spellings and decimal, octal and hex IPv4. validateNavigationUrl, the A/AAAA answer check and session-persist's cookie-domain guard all use it. BrowserManager now checks every navigation request (redirect hops, link clicks, scripted navigations): a blocked target resets the tab to about:blank and the write command fails with the reason. Connection-time DNS rebinding remains out of scope. Co-authored-by: Gabriel Rondon <grondon@gmail.com> * C4: external hosts point at carved sections relative to the installed skill Publishes the pointer form setup matches: on every non-Claude host a carved section is `sections/<file>` relative to the installed gstack-<skill> SKILL.md directory (QA's existing form), never $GSTACK_ROOT/<skill>/sections/, which does not exist in external runtime roots. No render changes yet: only QA is carved on external hosts today. * INV-1: gate outcome contract for outside reviews classifyOutsideReview() returns separate execution, findings and verdict fields from text, stderr and exit status; reason codes, anchors and fix actions live in lib/gate-outcomes.ts. validateOutsideReview() stays as the text-only wrapper and the two-argument CLI keeps exit 0/1/2 for importers and installed skills that predate a re-render. The --verdict form (implied by --stderr/--exit) prints VERDICT/FINDINGS/REASON and exits 0/3/4/1; the generated outside-voice caller moves to it and branches on each status (unverified prints OUTSIDE_STATUS: unverified, never completed). Adds docs/troubleshooting.md with one anchored section per reason code, the anchor test, and the gate-tail lint (query scripts may not swallow failures, gate calls may not end in || true, other best-effort tails say why). * G1: quote the base placeholder in document-release title sync The title-sync block is checked with bash -n; an unquoted <base> parses as a redirection. Migrates the test that pinned the old root VERSION read to the shared classifier call. * G3: give every logging-bin mktemp a ${TMPDIR:-/tmp} template Bare mktemp, mktemp -t and literal /tmp templates in the learnings, question and preference logs (plus jsonl-merge, distill, the two dashboards, /ship's PR body and /document-release's run dir) now use "${TMPDIR:-/tmp}/<name>.XXXXXX", the form the INV-3 lint accepts. When the temp file cannot be created, the learnings, question and preference logs now say the entry was not recorded and how to fix TMPDIR instead of exiting with only mktemp's own error. Three tests that pinned the bare PR_BODY_FILE=$(mktemp) and one that pinned the /tmp run dir migrate. GNU and BSD mktemp already honor TMPDIR without a template, so the macOS-sandbox drop in #2952 is unverified here; the template makes the temp location explicit on every mktemp. Co-authored-by: Davon Johnson <d.johnson@dayofshow.net> * B1b: structured outside-review gate fails on P0 as well as P1 validateOutsideReview() derived its structured pass/fail from a [P1] match, so a review reporting only [P0] findings passed. The wrapper now reads the classifier's findings verdict (P0 or P1), and the /review and /ship structured-gate prose requests [P0]-[P3] tags and fails on P0/P1; untagged output (OUTSIDE_STATUS: unverified) is missing coverage. The ship clarity pin moves from 'structured P1' to 'structured P0/P1' with the behavior. * E1: freeze normalizes boundary and target the same way on Windows; hooks spawn through cmd.exe (#2876, #2354) gstack_hook_normalize_path (careful/bin/hook-extract.sh) maps C:\..., c:/..., UNC \\server\share (leading // kept) and /cygdrive/c to one form; check-freeze runs both the saved boundary and file_path through it before the absolute test, after symlink resolution, and compares case-insensitively on Windows bash. freeze, guard and careful register hooks as bash -c "exec bash \"$HOME/...\"" so cmd.exe, which never expands $HOME, no longer turns the hook into a silent exit 127. Co-authored-by: Kevin Grasso <kevin@biblingo.com> Co-authored-by: Xuan-Ce Wang <261591073+xwang4-svg@users.noreply.github.com> * G4: keep CLI-to-daemon loopback calls off HTTP_PROXY (#494) The CLI appends 127.0.0.1, localhost and ::1 to the user's NO_PROXY (and no_proxy), keeping existing entries, before its first fetch. Verified on Bun 1.4.0: fetch honors NO_PROXY when it is set before the first request and ignores later changes, so it is set at the top of main(). * F1: bind the plan audit to this branch's plan, never the newest file Plan File Discovery now prefers an explicit binding (the plan-mode file or /autoplan ACTIVE_PLAN from the conversation, then a "Plan: <path>" line in the PR body), then lists candidates: repo-committed docs/designs/ files this branch changed or that name it, and personal plan files that name the branch. Candidates are offered, never picked silently. The repo-name match and the newest-file-of-the-day fallback, which audited PRs against unrelated plans, are gone. With nothing bound or chosen it prints the exact line 'Plan completion audit: not run (...). Fix: add "Plan: <path>" to the PR body, or run /autoplan.' /ship's parent now binds the plan before dispatching the audit subagent ({{PLAN_COMPLETION_GATE_SHIP:discovery}}), so the child never searches. Migrates three tests that pinned the old markers and notice. Co-authored-by: loulanyue <260355617@qq.com> * G5: browse type --selector <sel> <text> (#2936) --selector focuses the element (sel or @ref) and types into it. The bare form still types its text literally into the focused element; text that starts with -- goes after a -- separator. When the first word looks like a CSS selector, the result adds a hint to use --selector without echoing the typed text, and never refuses based on page contents. Help and command usage updated. Co-authored-by: Thomas Reuser <thomasreuser@gmail.com> * F2: design-doc discovery reads docs/designs, never a root DESIGN.md A root DESIGN.md is the design system /design-consultation writes, so preferring it unconditionally reviewed plans against a token file while printing "Design doc found" and never looked at docs/designs/ feature docs. Discovery now considers only docs/designs/*.md as the repo-local design doc (still freshness-compared with the private ~/.gstack copy). The CEO finding fixture seeds its supplied design as a docs/designs doc and its test follows. Co-authored-by: Bit0ps <alex.dreamscaper@gmail.com> * C4: rename the external carve test (typecheck:test union-order artifact) tsc prints a union's members in type-creation order; a test file sorting before ceo-mode-option.test.ts reordered one baseline diagnostic's text. * C1: every env-var host fence resolves its own gstack root Codex and the other env-var hosts run each block in a fresh shell, so GSTACK_ROOT/BIN, B and D set by the preamble were empty in later blocks (/bin/gstack-slug, B=/browse). scripts/resolvers/runtime-root.ts now owns root resolution: a valid exported GSTACK_ROOT (bin/ and lib/), else the per-install literal root (no git call), else repo-local, else the global root (CODEX_HOME on Codex); no root prints the setup fix line and exits. One post-render pass inserts it (<=400 bytes) into every fence that uses these variables without assigning them; B/D come from the same helper as BROWSE and DESIGN SETUP. Claude output is byte-identical. Migrated pins of the old preamble lines (gen-skill-docs.test.ts) and a fixture whose exported GSTACK_ROOT lacked lib/ (skill-positional-literals). * E8: headless browse honors GSTACK_CHROMIUM_PATH; extensions load in new headless mode (#2771, #1968, #2281) The headless launch passes GSTACK_CHROMIUM_PATH as executablePath, scopes the XProtect self-heal out for that bundle, and names the path in a launch failure with the fix. BROWSE_EXTENSIONS_DIR now launches a persistent context under the project state dir with --headless=new, so the extension runs and no window appears (the off-screen window was visible on macOS and the extension never ran in an isolated newContext). The GPU source guard test moves from the removed useHeadless flag to launch() vs launchHeaded(). Co-authored-by: oli548 <312288720+oli548@users.noreply.github.com> Co-authored-by: Slava <slava.istomin@tuta.io> Co-authored-by: ryanlaiwy <129729476+ryanlaiwy@users.noreply.github.com> Co-authored-by: 李翊楠 <yinanli1917@gmail.com> * C2: link design/dist and make-pdf/dist into every host runtime root and the agents sidecar Codex design and make-pdf skills fall back to $GSTACK_DESIGN and $GSTACK_MAKE_PDF under the runtime root, but setup linked only browse/dist, so both preflights reported NOT_AVAILABLE on a fresh install. One helper links dist/ only (make-pdf/ ships a Claude SKILL.md) into the Codex, Factory, OpenCode, Cursor, Copilot and Kiro roots, the .agents sidecar and the Cursor sidecar. Co-authored-by: Mammad M. <mammad@digit.az> * C3: install router and nested SKILL.md files as real copies; back up hand edits; --status checks Codex skips a symlinked SKILL.md file, so the gstack router linked into ~/.codex/skills/gstack was invisible. Every host runtime root (Codex, Factory, OpenCode, Cursor, Copilot, Kiro) and Kiro's per-skill installs now get real copies, rewritten on every run (the auto-update path runs setup --refresh-registered, so it refreshes them too). setup records each copy's hash in <state root>/skill-copies.tsv; a copy that no longer matches is saved to a uniquely named backup under <state root>/backups/skill-copies/ before _activate_runtime_root replaces the tree, and a failed backup keeps the old tree. setup prints what it refreshed or removed. ./setup --status now prints, per registered install, whether the router is a real file and whether every section pointer resolves, with the fix command. Test migration: setup-codex-model pinned the Kiro router/upgrade links as _link_or_copy; C3 deliberately changes them to copies. * INV-3: install-level host conformance checks (no symlinked SKILL.md, dist present, section pointers resolve) For codex, kiro, factory, opencode, cursor and copilot, the real-setup conformance kit now fails when an installed SKILL.md is a symlink, when design/dist or make-pdf/dist is missing from the runtime root, or when a 'Read `.../sections/<f>.md`' pointer in an installed SKILL.md does not resolve (bare pointers in carved skills, relative-to-installed-skill pointers, and $GSTACK_ROOT/~/absolute/../ pointers everywhere), and ./setup --status must report 'all resolve'. No new setup runs. Expected red until the generator lane lands: design-html's install-offer pointer and plan-eng-review's review-sections pointer name $GSTACK_ROOT/<skill>/sections/... on every external host. * B11: auto-update tracks pull, setup and migrations; announces only on success; retries with backoff gstack-session-update wrote the 'just upgraded' marker after SETUP_FAILED or SETUP_SKIPPED bun_missing, and re-ran setup only when HEAD moved, so a failed install was announced as an upgrade and never retried. Stages are now persisted in <state root>/session-update-pending; the marker is written only when every stage succeeded; a pending setup or migrations stage is retried even with HEAD unchanged after the update-check interval, then 6h, then 24h; and each session start prints one line naming the stage, the reason and the repair command while it is pending. Best-effort tails carry a same-line reason. * D1: gstack-config key browse_extension_id for forks and self-built extensions The browse daemon trusts one Chrome extension ID on /extension-token and the terminal-agent /ws. The override now lives only in gstack-config (never an env var a project .env could set): empty by default (gstack's published extension), a 32-letter a-p Chrome ID when set, "" clears it, and anything else is rejected with the stored value kept. The browse lane reads it. gstack-config-defaults: browse/src counts as a reader directory; the key is listed prose-only until the browse lane's reader lands (the stale-entry check then asks for the entry's removal). * G2: key the sticky slug cache on the git common dir A linked worktree used to resolve its slug fresh from the remote while the main checkout kept its sticky pre-remote identity, so learnings, checkpoints and timelines split into two buckets. gstack-slug and its TS twin in lib/bin-context.ts now key the cache on the git common dir (read from .git files, no git spawn, so cache hits stay git-free) and read entries in one deterministic order: the repository entry, then the main checkout's pre-upgrade entry, then this directory's own. Both the repository and per-path entries are written. When a worktree's own old entry named a different bucket that still holds data, that slug is recorded in projects/<slug>/.former-slugs and reported by --get FORMER_SLUGS. New --adopt-legacy --from <slug> merges it: JSONL logs append missing lines with de-duplication, missing files copy, differing files are listed as conflicts and never overwritten, the source bucket is kept, and it prints migrated/pending/conflicting counts; reruns are idempotent. --adopt-legacy --dismiss <slug> drops the reminder. Co-authored-by: loulanyue <260355617@qq.com> * C5: OpenCode /gstack-* commands load the skill by its frontmatter name OpenCode's skill tool knows gstack-review as `review` (its frontmatter name), and OpenCode's builtin /review shadows that name as a command. The managed /gstack-review command told the agent to load `gstack-review`, a name the skill tool does not list. Every managed command now names the frontmatter name. OpenCode's AskUserQuestion/ExitPlanMode mapping (#2759) already shipped in v1.91.16.0. Test migration: the #2629 commands test pinned the directory name. Co-authored-by: Lockyer <135391289+Lockyer228@users.noreply.github.com> * E4: Windows node-server bundle keeps sharp and socks external; CLI resolves its server script lazily (#2260, #2439) build-node-server.sh passes --external sharp and --external socks (Bun on Windows refuses to inline them; on Linux they were silently inlined). The CLI no longer resolves server.ts or server-node.mjs at module load: resolveServerLaunch() runs when a server starts, so a Windows minimal runtime with only server-node.mjs works and a missing script is reported only when one is needed. Co-authored-by: Spooks444 <193448029+Spooks444@users.noreply.github.com> Co-authored-by: punksterlabs <77217235+punksterlabs@users.noreply.github.com> * B1: Codex reviews whose sandbox could not start are unavailable, not a pass In containers without unprivileged user namespaces every Codex command fails, yet the review exits 0 and concludes 'no findings' (captured in test/fixtures/codex-sandbox/). The validator now takes execution evidence: 'codex exec --json' command events and the 'codex review' stderr transcript (a command that succeeded proves execution); sandbox failure lines on stderr or in failed command output return unavailable(sandbox_unavailable); exact 'could not run commands / read the diff' phrases are the last fallback. Review text that merely discusses bwrap or namespaces still passes. A free 'codex sandbox ... true' preflight (Linux, 10 s) runs before the paid model probe and every automatic outside review; CODEX_MODE gains sandbox_unavailable, and an inconclusive model probe reports unverified instead of ready (exit code kept for stale callers). Every Codex site (outside-voice resolver, /codex review, challenge, consult) takes its sandbox from _GSTACK_CODEX_SANDBOX and pipes its result through the validator; a static test parses rendered command positions for both. GSTACK_CODEX_NO_SANDBOX=1 (exactly 1, shell env only) switches to danger-full-access with a warning on every use. Sandbox pins in existing tests move from the literal read-only to the selected-sandbox variable; captured autoplan probes are judged against the probe block they executed. * C6: gstack-relink discovers skills structurally, so /browse survives a skill_prefix toggle relink skipped a hardcoded list of support-directory names that included browse, a real skill, so toggling skill_prefix removed /browse. A top-level directory is now a skill exactly when it carries a SKILL.md. Co-authored-by: Jonathan Zhang <topcoder1@Mac-Studio.local> * C4: carve ship and plan-ceo-review on every host; enforce the 160 KB byte gate The generated ship skill was 202-241 KB on all 10 non-Claude hosts (and Factory plan-ceo-review 160,039 bytes), over the 160,000-byte ceiling, because section carving ran only for Claude. Both skills now carve on every host, pointing at sections relative to the installed skill directory (literal $GSTACK_ROOT/<skill>/sections refs are rewritten the same way). The ceiling is measured in UTF-8 bytes and fails generation in repo mode; under --install-root it warns so setup never leaves a user without skills. qa-lazy-sections.test.ts pinned ship as inline-on-external-hosts; migrated to the carved policy and kept the inline check on review (Claude-only carve). * B5: learnings and timeline reads fail loudly when bun is missing gstack-learnings-search and gstack-timeline-read ended in '2>/dev/null || exit 0', so a missing bun or a crashed embedded script printed nothing and exited 0, identical to 'nothing recorded'. Both now exit 127 with '<script>: bun not found on PATH, so ... could not be read. Fix: install Bun (https://bun.sh), then re-run ./setup.' and let an embedded-script failure exit non-zero with its stderr. A successful run with no matches still exits 0 with empty output. The generated callers that print 'LEARNINGS: unavailable (...)' belong to the generator lane. Co-authored-by: loulanyue <260355617@qq.com> * B2: a retired-model 404 reports MODEL_UNUSABLE, not CODEX_MODE: ready The model probe matched only 400 / 'is not supported', so a retired model's 404 ('does not exist or you do not have access to it') and a custom provider's bare 404 fell through to the inconclusive branch and reported ready. Both now classify as MODEL_UNUSABLE with a hint naming the two causes (retired model, or a wrong base_url) and log codex_model_retired; a 'requires a newer version of Codex' 400 points at upgrading the CLI and logs codex_cli_outdated. Ported from #2843 onto the current probe. Co-authored-by: Marek Cermak <marek.cermak@strv.com> * B6: rank learnings by matched query tokens and disclose truncation A more specific query used to drop the exact match: the token-OR filter recalled every entry sharing one token, the sort used confidence alone, and the default --limit cut silently. Entries now rank by how many distinct query tokens they match, then confidence, then recency, and a query that matched more than fit says '-- N more matched, raise --limit to see them'. The hit count lives in a local map so a field stored on disk can never become a sort key, and the type grouping uses a null-prototype map so a row typed 'constructor' or '__proto__' cannot crash the run. Ranking code ported from #2796; regression vectors adapted from #2799 (its stemmer is not needed). Co-authored-by: loulanyue <260355617@qq.com> Co-authored-by: y$un_ <forrest.sun527@gmail.com> * C7: required team mode also blocks Copilot CLI's lowercase skill tool Copilot CLI applies Claude-format matchers by tool name; its skill tool is `skill`, which no Claude tool name maps to, so the 'Skill' matcher never fired and Copilot ran protected skills without gstack. The hook entry now matches 'Skill|skill', an entry an older team-init wrote is upgraded in place, and the deny is emitted in both decision formats (Copilot reads the top-level permissionDecision, Claude Code reads hookSpecificOutput) with exit 0 so both parse it. Test migration: the settings test pinned matcher 'Skill'. Co-authored-by: neallee <neallee@gmail.com> * B3: Codex auth probe accepts custom providers' env_key credentials The auth probe only knew CODEX_API_KEY, OPENAI_API_KEY and auth.json, so a working self-hosted or OpenAI-compatible provider configured with [model_providers.<id>] env_key = "X" reported AUTH_FAILED and /codex stopped. The probe now reads env_key from model_providers tables only (comments ignored, the name must be a shell identifier before it is dereferenced) and accepts a set variable; when none is set it names the missing key on stderr. Ported from #2192 with the table scoping added. Co-authored-by: hardy <932821449@qq.com> * E8: keep the extensions profile under the state dir at a path lock cleanup accepts cleanSingletonLocks only cleans directories named chromium-profile, so the BROWSE_EXTENSIONS_DIR profile moves to <state>/extensions/chromium-profile. * B10: plan-tune never reports CALIBRATED with zero signal events Stats printed CALIBRATED: true from logged-answer counts alone, while every answer could miss SIGNAL_MAP and leave each dimension at its 0.5 seed (#2623). --derive now records inferred.signal_events, the answers that actually moved a dimension; calibration requires at least one, and the summary prints 'not calibrated: no recorded signals' verbatim when there are none (or says the count is unknown for a profile derived before this change). The psychographic redesign stays out of scope. * A1: verify gbrain landing before stamping transcripts, plus --reconcile gbrain import exits 0 while skipping pages (returned per-file failures, malformed filenames, managed-import Pending receipts), and its failure ledger exists only for git import dirs, so a skipped page was stamped ingested and never retried (field reports: 37 and 69 lost entries). Write, verify, then stamp (lib/memory-ingest-landing.ts, a new ratcheted module): - State schema 2 with a transition table (staged -> imported_unverified -> ingested, plus refused, quarantined, unrecoverable). Schema 1 files migrate in place: entries become ingested in the legacy "default" source and a reconcile pass is marked pending. - Named failures come from --json `failures`, stderr "Skipped <rel>:" lines and the ledger; a failure gbrain does not attribute refuses the whole batch. After 3 refused batches the batch is bisected and the page that poisons it is quarantined (printed even under --quiet). - Landing check: one `gbrain list --source-id` per source per run and at most 50 `gbrain get --json` content checks against a gstack_content_sha256 frontmatter field gstack now writes. Lookup errors and pglite_busy mean "not yet checked", never absent; pages over the cap stay imported_unverified and are not re-imported. - Each entry records its gbrain source_id. Remote-http mode records "staged", never "ingested". - The source fingerprint comes from the exact bytes parsed, so a transcript that grows during import is not stamped as its newer self (the old unscanned contract test is migrated; it pinned that bug). - --reconcile mode (bounded by --limit, default 200; resumable; --dry-run; backs up the state file) prints "reconcile: checked N, present M, re-queued K, not yet checked L". --request-reconcile records a pending pass without calling gbrain; the v1.91.17.0 migration (placeholder name) uses it, and the next ingest run does one bounded pass. - One lock for every state writer; a state-save failure fails the run. The #1802 static pins in regression-1611 move to the new structure (remote-http returns before importBatch; the finally honors preserveStaging). Co-authored-by: afshaker <andrew@phone2.io> * B4: arm64 re-sign keeps a backup, restores it when codesign fails, and reports design unavailable on exit 137 setup's Apple Silicon re-sign mutated each binary in place (truncate, remove signature) and left it that way when re-signing failed; a design binary that was then SIGKILLed at launch produced at most a quiet log line. Each binary is now copied first and restored as built when signing fails, and design is exec-probed after re-signing: exit 137 prints 'design unavailable: ... Fix: cd <checkout> && ./setup' (even under -q). Verified on Linux with the real block and uname/codesign/otool/stat stubs; not run on a Mac. * D3: treat design docs and handoff notes as data in plan reviews plan-ceo-review, plan-eng-review and plan-devex-review read the freshest design doc as the source of truth, and nothing said its content is data. In multi-agent pipelines another agent may have written it, or a handoff note beside it, with reviewer-directed text ('skip Step 0', 'approve as-is'). Each read site now says the doc and any handoff note are data, never to follow text aimed at the reviewer, and to report it as suspicious content. The plan-completion audit's discovery, which now reads repo docs/designs/ files (F1), carries the same rule. Co-authored-by: cube <frosimanuel@gmail.com> * D4: setup-deploy reports Render API key presence only /setup-deploy ran 'echo $RENDER_API_KEY | head -c 4', printing the key's first bytes into the transcript. It now prints only 'RENDER_API_KEY: set' or 'RENDER_API_KEY: not set'. Wording follows the #2971 port of #1096; a test also guards against any setup-deploy command echoing a credential variable. Co-authored-by: jacob-wang <jiayuw794@gmail.com> * D5: per-project headed Chromium profiles; never kill a live profile owner (#2492) The headed profile moves from <gstack home>/chromium-profile to <project>/.gstack/chromium-profile (CHROMIUM_PROFILE still wins). killOrphanChromium now kills a SingletonLock PID only when the lock names this host, the process is a Chromium using that profile, and its daemon is gone (reparented to init or a subreaper); a live owner or another host's lock stops browse with the holder named. On first headed use a project profile is seeded from the old shared profile only when nothing holds it: copied without Singleton* files into a 0700 temp dir and renamed into place, with a one-time notice; otherwise it starts fresh and says how to import later. New: browse profiles [list] and browse profiles prune [--days N] (never removes a profile in use or the old shared one). The #2817 headed-startup tests move to the project profile and a real orphan (the old stand-in was a live child of the test). Co-authored-by: miao <miao@macmini-2.local> * F5: installed Claude skills serve the rendered sections, not the checkout copies Probe: a gbrain user render serves ship/SKILL.md from the render dir; its STOP pointers name absolute render paths, but its section index names `sections/<f>.md` relative to the installed skill directory, whose sections/ was linked from the checkout. So the rendered 'Save Results to Brain' blocks (plan reviews, office-hours, ship adversarial) were generated and never read. setup's _link_skill_runtime_assets and gstack-relink (the gbrain-refresh path) now serve a rendered counterpart per file, keep checkout-only assets (manifest.json), return an asset to the checkout once its render is gone, and never touch an entry that is not gstack's links. Co-authored-by: Jonah Berg <jonah.berg.g@gmail.com> * F1: ratchet /ship's parity union-size cap for this lane's content G1's NO_VERSION arms, F1's parent-side plan binding and D3's data line add 2,126 bytes to /ship's skeleton-plus-sections union, taking it from 1.4022x to 1.4135x of the v1.64.1.0 baseline (cap was 1.405). The cap moves to 1.414 with the measurement recorded, as earlier growth did; no other invariant changes. * E5: feed every Codex prompt on stdin instead of argv Prompts went to Codex as one argv string ('codex exec "$(cat file)"'), which fails over 32 KB on Windows (and over 128 KB on Linux, E2BIG) and mangles quotes. Every codex exec site (outside-voice resolver, /codex review custom instructions, challenge, consult new and resume) now runs 'codex exec -' with the prepared prompt file on stdin; challenge and consult write the prompt with a quoted heredoc so it is never shell-parsed. The bash-native timeout watchdog (stock macOS) started the command in the background, which gives it /dev/null as stdin; it now passes the caller's stdin through. The 'byte for byte' dispatch test drops its old argv trailing-LF exception, the consultation and autoplan fixtures follow the stdin form, and the static site test requires stdin at every rendered codex exec. * G6: prettyscreenshot --hide keeps a trailing image path as the output path (#1419) --hide collected every following non-flag argument as a selector, so the output path was hidden instead of written to. A trailing argument that looks like a path or ends in .png/.jpg/.jpeg/.webp is now the output path. Co-authored-by: Jayesh Betala <jayesh.betala7@gmail.com> * INV-3: render-level host conformance and generated-bash lint on every host One render per host checks every fence of every SKILL.md and section file: bash -n with angle-bracket placeholders normalized; the env-var host prelude and its B/D consumers run under env -i and set -u with stub binaries (no empty value, no /bin/, /browse, /design or /gstack- path); and a quote-aware lint (test/helpers/generated-bash-lint.ts) fails on a ~/ word inside double quotes, mktemp without a TMPDIR-respecting template, and cd/rm -r/mv/git -C on a variable the fence never assigned unless written ${VAR:?...} (cd must also handle failure). Negative controls plant each failure. Templated the bare mktemp calls in generator-owned resolvers and the upgrade fence; migrated the two tests that pinned the bare form. * G6: analytics averages quoted duration_s values (#1728) gstack-skill-end writes duration_s as a quoted string, and the per-skill average stripped everything from the leading quote, so every average read 0s. The awk now drops a leading quote before extracting digits. Co-authored-by: Ryan Alberts <ryan.a.alberts@gmail.com> * G6: gstack-ios-qa-daemon --help prints usage instead of starting the daemon (#1932) Co-authored-by: maxpetrusenkoagent <max.petrusenko.agent@gmail.com> * G1: tighten /ship's NO_VERSION wording to fit the union-size cap Same behavior in fewer bytes: the NO_VERSION arm, the broken-source exit, the unprefixed-title rule, the Step 20 null and the CHANGELOG skip. The cap ratchet commit measured this wording. * G6: rejected taste preferences gain confidence (#1777) bumpPref computed every entry's confidence from approved_count, so a rejected entry, which never gains approvals, stayed at 0: show ranked rejections by insertion order and the taste-drift warning could not fire from real use. Confidence now uses the entry's own bucket count. Also salvaged by Garry's #2971. Co-authored-by: Jayesh Betala <jayesh.betala7@gmail.com> * E2: Windows /cso build disables only C5105, and a failed MSVC compile is not reported as a missing toolchain The Windows SDK's winbase.h raises C5105, which /W4 /WX promotes to error C2220, so every MSVC build of the native /cso helper failed, and setup mapped every probe failure to 'install Visual Studio 2022 Build Tools'. Both compile lines add /wd5105 (only that diagnostic), and setup keeps the probe's output: a probe that found MSVC but failed to compile reports windows-msvc-compile with the compiler's first error. Not run on Windows: the mapping test stubs powershell.exe; the ps1 assertion runs everywhere. Co-authored-by: AlinValentin7 <AlinValentin7@users.noreply.github.com> * E6: gbrain-refresh hands the detection file to Python on stdin Under Git Bash, a native Windows Python first on PATH cannot open an MSYS /c/... path, so gbrain-refresh always summarized the detection as 'local-status: unknown' and brain-aware blocks were suppressed. bash now opens the file and Python reads stdin; the grep fallback without Python is unchanged. Verified on Linux with a stub Python that refuses path opens; not run on Windows. Co-authored-by: maxpetrusenkoagent <max.petrusenko.agent@gmail.com> * G6: context-save computes session duration on Linux (#2704) The duration step parsed 'ps -o lstart=' with BSD-only 'date -jf', so on Linux session_duration_s was always unknown. It now falls back to GNU 'date -d', and never parses an empty start (GNU reads that as midnight). Ported onto the current block instead of #2704's ~/.gstack/sessions mtime, which bypassed the state root. Co-authored-by: Evgenii Lopatin <e75533@gmail.com> * Q2: one-time notice naming the Codex provider and account for outside reviews Outside reviews are on by default and use whatever Codex login exists (#965). The first automatic Codex review on a machine now prints one non-blocking notice: the provider (config.toml model_provider, else openai), where the credential comes from (CODEX_API_KEY, OPENAI_API_KEY, the ChatGPT login or the key saved by codex login; never the value), that reviews stay on, and how to turn them off (gstack-config set codex_reviews disabled). A marker in the gstack state root keeps it to once per machine. Tests that run the generated command now use private state roots. Also shortens the no-sandbox provenance line. * Q1: validate browser output paths the way the kernel resolves them (#2971 salvage) validateOutputPath walks the destination one component at a time, resolving each existing component before the next, so link/../x, a dangling symlink and a dangling symlinked parent can no longer route a write outside the safe directories, while a new multi-level directory inside them is allowed. scrape validates its manifest and every downloaded file, not only --dir. Ported from PR #2971 onto current code. * G6: retro test census counts non-JS suites, skips generated output The census recognized only JS/TS-style names, so pytest and minitest (test_*.py|rb), XCTest (Tests/ dirs, FooTests.swift), Terraform (*.tftest.hcl) and Bats (*.bats) suites counted as zero, and a committed build directory counted as authored code (and its Test/ subdirs as test LOC). is_test now knows those conventions and never treats prose as a test, generated output is skipped in the numstat pass, TEST_FILES_CHANGED uses is_test, and TEST_FILES_TOTAL uses the same filename conventions minus generated paths. Merges #2812, #2809 and #2037's patterns (#2037's retro template edits target a block the metrics bin replaced). Co-authored-by: Jino <dgfrty00@gmail.com> Co-authored-by: snig-17 <santosh.m.tiwari@gmail.com> Co-authored-by: Sholto McNeilage <sholto@nodenarrative.com.au> * Q3: setup stops auto-installing the AskUserQuestion preference hook in Conductor and removes one it added The PreToolUse question-preference hook breaks Conductor's native AskUserQuestion round-trip even when it only defers (#2207), and setup auto-installed it on every Conductor workspace setup. Now a Conductor setup without an explicit opt-in never installs it and removes an existing one (the question-log and AUQ-error hooks stay), printing how to keep it (gstack-config set plan_tune_hooks yes). An explicit opt-in still installs it. gstack-settings-hook remove-source gains --event to scope the removal. Test migration: setup-plan-tune-hooks-noninteractive pinned the retired Conductor auto-opt-in. Co-authored-by: Phil Hie <philhie@outlook.com> * E5: watchdog stdin test feeds the wrapper through a redirect Callers run '_gstack_codex_timeout_wrapper ... codex exec - < file'. Bash keeps stdin for a backgrounded command inside a pipeline but not behind a redirect, so the pipe form passed on the old watchdog; the redirect form reproduces the lost prompt. * F4: type the variant-round fetch stub like the existing stubs The test-typecheck ratchet flagged the zero-argument stub cast to typeof fetch; it now takes the fetch arguments like variants-retry-after's stub. Test behavior unchanged. * A4: import transcripts into per-repository, non-federated gbrain sources `gbrain import` ran without --source-id, so transcripts from every project landed in the shared (often federated) default source. - Each transcript partition imports with --source-id into its originating repository's own source (gstack-transcripts-<repo>, built with constrainSourceId, now exported from lib/gbrain-sources.ts). The source is registered machine-local and --no-federated before its first import, and the registration is cached in state. Partitions import one after another (PGLite is a single writer); a failing source no longer blocks the others, except a timeout, whose checkpoint must survive. - If gbrain cannot register the source, that partition stays local and unstamped, and the run says why (even under --quiet). - --include-unattributed sends repo-less transcripts only to the machine-local gstack-transcripts-unattributed source, keeps them on the machine when the brain is remote (Postgres, Supabase or remote-HTTP), and never writes them to the publishable ~/.gstack/transcripts staging. - A page's recorded source is sticky: pages stamped before this release ("default") keep importing there, so no source gains a duplicate. - The staging dir name carries -src-<id> so a resumed import keeps its source. Test fakes that recorded only the last import now record every import of one run, and answer `gbrain sources`; remote-http scan tests use an attributed transcript, since unattributed ones no longer stage there. Co-authored-by: Sinabina <sinabina@Sinabinas-MacBook-Pro-4.local> * F3: load the taste profile, and report a failed slug lookup The taste-profile probe double-quoted ~/.claude/.../gstack-slug, so the tilde never expanded and /design-consultation and /design-shotgun always printed NO_TASTE_PROFILE. Use quoteSafePath(); stop hiding gstack-slug's stderr and print TASTE_PROFILE_UNAVAILABLE with the fix when the slug cannot be resolved. * B4: DESIGN_READY only when the design binary actually starts DESIGN SETUP (and office-hours' visual exploration) printed DESIGN_READY for any executable binary, so one SIGKILLed at launch by an invalid macOS code signature still passed. Launch it under the portable 10-second deadline (gtimeout, timeout, or perl alarm); otherwise print DESIGN_NOT_AVAILABLE with the exit code, 'timed out after 10s', or on 137 the code-signature hint and 'Fix: cd <checkout> && ./setup'. A success is cached by inode and mtime under the state root so Gatekeeper's first-launch scan is paid once. The probe launches the binary with no arguments (usage, exit 0): the design CLI has no --version command today. * C2: name the runtime root every way a generated preflight may read it in the dist test The generator lane's C1 resolves D from $GSTACK_ROOT instead of $GSTACK_DESIGN, so the test now exports GSTACK_ROOT as well; it still fails on main (no dist in the roots) and passes with and without the C1 change. * A5: brain-sync stage fails loud when the artifacts source indexed zero pages The stage reported "curated artifacts pushed" whenever gstack-brain-sync exited 0, but a git push is not gbrain indexing. After a successful push the stage now reads the maintained artifacts source's page count (GSTACK_BRAIN_SOURCE_ID, else the source whose path is the artifacts worktree or state root): zero pages is ERR with the fix command; a missing source or an unavailable count says so instead of claiming indexing. Co-authored-by: Jimmy Zeisweiss <4261258+jzeisweiss@users.noreply.github.com> * B1: state the effective read-only sandbox in /codex review prose The codex parity invariant requires the skill to state sandbox_mode="read-only"; the command now takes it from _GSTACK_CODEX_SANDBOX, so the prose names the effective default explicitly. * A6: report deleted-worktree gbrain sources; prune only with proof, on request A deleted Conductor worktree leaves its gstack-code-<hash> source behind and gbrain's autopilot fails on it forever. Local absence is not proof a source is dead (machines sharing a federated brain register their own paths, and its pages may be the only copy of uncommitted work), so gstack never removes one automatically. - Each sync counts consecutive misses for gstack code sources whose path is missing, prints "path unavailable" once per source with the review command, and records which repository each synced source belongs to. - New gstack-gbrain-sync --prune-gone-worktrees (with --dry-run, whose plan it always prints first) removes a source only when its path was missing on two consecutive syncs, its parent is readable, recomputing deriveCodeSourceId() for that path on this host yields its id, and the repository's git common dir still exists and no longer lists the worktree. It re-checks the path right before removing (through the #1734-guarded safeSourcesRemove) and logs each removal. * D0: pass the Bun child flags after 'run' so wrappers that rewrite 'bun run' argv keep working bun run --no-env-file --config=/dev/null <script> has the same effect (verified on Bun 1.4.0) and keeps argv[1] === 'run', which the publication-lock fixture's preload injection relies on. The sidebar-tabs source guard now expects the flags in the terminal-agent spawn. * E8: count the BROWSE_EXTENSIONS_DIR persistent launch in the signal-flags guard The new launch site carries handleSIGINT/SIGTERM/SIGHUP false like the other three, so the guard's expected count moves from 3 to 4. * D2: give the memory-leak reproducer's fake request isNavigationRequest() wirePageEvents now runs the navigation guard on every request event; the fake request adapter supplies the method real Playwright requests have. * A7: run only the resolve_symbol_edges dream phase, detected first /sync-gbrain --dream and the --full auto-build ran gbrain's full dream cycle (~35 minutes, LLM phases) when the call graph needs only the resolve_symbol_edges phase. The stage now runs gbrain dream [--source <id>] --phase resolve_symbol_edges when the installed gbrain accepts that phase (detected by behavior: gbrain validates --phase before printing --help). Otherwise it skips the forced dream and says the full cycle costs about 35 minutes and how to run it, instead of starting it. Co-authored-by: liutiming <timingliu@outlook.com> * D0: keep the dotenv isolation test runnable on the Windows lane The compiled probe gets its .exe name on Windows, and the terminal-agent case (owner identity comes from ps) is skipped there. * B9: read cycle_freshness from a doctor run that includes it /sync-gbrain Step 3.5 and the --full auto-dream decision read CYCLE from gbrain doctor --json --fast, which skips the DB checks that carry cycle_freshness, so the answer was always "unknown". Both now read gbrain doctor --json --scope=brain (DB checks, no skill-file walk), parse its JSON even when doctor exits 1 for a failing check, and match the issue that names this source: "never completed" is never, "last cycled Nh ago" is completed. When the installed gbrain does not report the check, they say "unknown: installed gbrain does not expose cycle_freshness" (lib readCycleStatus carries the reason to the --full SKIP row). The source-probe fixture that pinned the --fast call and a made-up message format now answers --scope=brain with gbrain's real message. * B4: compact the launch probe (cache keyed on inode, valid while newer than the binary) Keeps design-consultation inside its parity size budget. * C4: migrate the ship docs-dispatch test to the carved external ship On Codex and Factory the documentation dispatch now lives in sections/documentation.md behind a pointer relative to the installed skill, like Claude; the test pinned the old inlined layout. * B5: learnings callers report a failed search instead of reading as empty Every generated gstack-learnings-search caller ended in 2>/dev/null || true, so a missing bun or a failed search looked like 'nothing recorded'. Callers now keep stdout, capture stderr and the exit status, and print 'LEARNINGS: unavailable (<first stderr line>)' on failure (resolver callers in learnings.ts and review-army.ts, plus investigate, qa and ship's adversarial section). * F1: ship and review count unverified/unavailable outside reviews as missing coverage INV-1 consumers in this lane's templates: /ship's readiness note, its PR body Pre-Landing Review section and /review's completion rule now say an outside review reporting unverified or unavailable is missing coverage, never a pass. The PR body's Plan Completion section prints F1's not-run line instead of the old 'No plan file detected.' The /ship union-size cap moves to 1.415 for these lines (measured 1.4148). * A8: rename the artifacts repo with OWNER/REPO, rewrite the remote only after it The v1.27.0.0 migration passed a bare repo name to gh repo rename / gh repo edit, which always fails (gh wants [HOST/]OWNER/REPO), then step 2 rewrote ~/.gstack-artifacts-remote.txt to the never-renamed repository. - Qualify both names with the owner from the recorded remote URL (else the signed-in gh user) and surface gh's error. - Step 2 moves the remote file but rewrites the URL only once the rename is confirmed; a later successful retry rewrites it then. - A pending journal now counts as state to migrate, so the #1383 retry actually re-runs step 1 (it used to see no legacy file and declare the migration done). - The v1.91.17.0 migration (placeholder name) repairs installs already hit: when the remote names gstack-artifacts-<user> that does not exist but gstack-brain-<user> does, it restores the old URL in the remote file and ~/.gstack's origin and says so. It only calls gh for such a remote and never fails the upgrade. Co-authored-by: Brandon Pugsley <aibpaem@gmail.com> * B8: read gbrain JSON with bun instead of jq in install and wireup gstack-gbrain-install read package.json with jq but never checked for it, so on a machine without jq a successful install failed with "cannot read version". gstack-gbrain-source-wireup hard-failed without jq ("brew install jq" is not actionable on git-bash), so the wireup was skipped on every run. Both now read the JSON with bun, which they already require. Co-authored-by: benjamin beres <benjamin.beres@bienpreter.com> * B4: shorter probe messages (design-consultation parity budget) * B5: shorter capture variable names (parity budget) * E7: find the Aside CLI in ~/.local/bin when it is not on PATH (#2902) The installer puts aside in ~/.local/bin, which is not on the default macOS login PATH, so the probe said NEEDS_ASIDE ('download Aside') for a healthy install. The probe now falls back to ~/.local/bin/aside, drives the resolved binary, and prints READY: <path> when it is off PATH so later blocks run that path. Ported from PR #2965 onto the current _gs_d probe. Co-authored-by: Som Samantray <som.samantray@gmail.com> * A2: key the dot-slug catalog entry on the message users actually saw * E1: cmd.exe-safe investigate freeze hook On Windows Claude Code runs hook commands through cmd.exe, which cannot parse the single-quoted bash -c argument, so /investigate's scope-lock hook never ran. Use one double-quoted bash -c script that names the hook path directly (no S= variable: POSIX sh would expand $S to empty and skip the hook). * F4: design resolver reads per-round variant paths $D variants now writes $_DESIGN_DIR/round-<N>/variant-*.png and prints JSON with roundDir, manifest and paths. The mockup flow and comparison-board loops build the board from those printed paths instead of hardcoded variant-{A,B,C}.png, and approved.json records approved_path and round so a round-1 choice made after round 2 still resolves. * E4: generators tolerate Bun's spurious EEXIST on Windows (#2329) Bun on Windows throws EEXIST from a recursive mkdir of an existing directory, so gen:skill-docs and the llms.txt writer crashed once a second file landed in a directory. Both now use lib/fs-utils' mkdirpSync (ported from PR #2329 onto the existing helper instead of a new scripts/fs-utils.ts). Co-authored-by: Vaughan <vaughan@cairnsitguy.com.au> * C7: keep the required-mode hook's exit-2 deny Copilot CLI treats a preToolUse exit 2 as deny (its hooks reference), and #2413 established exit 2 plus hookSpecificOutput as the shape Claude Code enforces, so the deny output stays as on main; only the matcher and the in-place upgrade of older 'Skill' entries change. * C3: guard Kiro's router copies on a present source; give the QA pointer adapter the copy and dist helpers A copy fails where a symlink dangled, so the Kiro router and upgrade copies are written only when their render exists (as every other runtime root already does). test/qa-lazy-sections' extracted Kiro installer now supplies _link_runtime_dists (C2) and the SKILL.md copy helpers (C3). * Q3: office-hours asks open-ended questions in a Q<N> prose form (#2719) In Conductor, and whenever AskUserQuestion fails, prose is the only channel, and the prose fallback defined only the D<N> decision brief. Phase 2A/2B diagnostic questions have no option set, so the model dropped the phase. Office-hours' preamble now defines the Q<N> open-question form (question, why I'm asking, what a strong answer sounds like) and routes free-text replies to it; the diagnostic sections name it. Scoped to office-hours so every other skill's preamble is unchanged. Ported from PR #2729. Co-authored-by: Bharadwaj Mittapelli <bharadwajmittapelli@gmail.com> * A2: render db-unreachable like timeout gbrain's new db-unreachable status is a transient network failure; a setup run while offline must not strip brain-aware blocks from every skill. * A2: treat db-unreachable like timeout in gbrain-refresh The memory lane's new db-unreachable status (network or DNS failure, config intact) is transient, so gbrain-refresh keeps rendering brain-aware blocks for it exactly as it does for timeout instead of suppressing them. * C8: a render given disabled skills drops them from the router (#3018) The root router named every skill and ends 'When in doubt, invoke the skill', so it routed to skills the user disabled. gen-skill-docs takes --disabled-skills (runGeneration disabledSkills): the router drops their routing rules, keeps the other side of an 'A or B' rule, and names them as disabled. The router and gstack-upgrade cannot be disabled. With no disabled skills the render is byte-identical. * G2/G3/INV-1: skill-start former-bucket notice and visible learnings failure; TMPDIR mktemp in gbrain bins; best-effort reasons * C9: Step 2 falls back to the install registry's source checkout (DX-7) Codex users who installed from the README's ~/gstack clone got 'ERROR: gstack not found' from /gstack-upgrade: Step 2 only looked under host skill directories. It now falls back to the source checkout the install registry recorded for the current host. * B4: design --version prints the version, and the probe uses it The design CLI treated --version as an unknown command (exit 1), so the probe launched the binary with no arguments instead. Add --version (prints the gstack version, exit 0, before auth or network) and probe with it, so messages read '<path> --version exited <code>' as planned. * C4: no external render points at a section file that is not installed design-html's detector install offer and plan-eng-review's self-check pointed at $GSTACK_ROOT/<skill>/sections/<file> on every external host, where neither skill is carved and no section tree exists. design-html now inlines the offer off Claude, and a skill's pointer to its own {{SECTION}}-inlined section names the inlined text. * C1: skillify finds the browse SDK under its own host's install root The template hardcoded ~/.claude/skills/gstack for the canonical browse-client.ts, so external hosts read another host's install or nothing. A new INSTALLED_ROOT resolver (sibling of INSTALLED_BIN_DIR) names the host's global root, or the per-install root when setup renders for one. * C4: rewrite section refs on hosts whose root stays a literal ~/<globalRoot> Kiro, OpenCode, Cursor and Copilot render ~/.claude/skills/gstack as their own literal root, not $GSTACK_ROOT, so plan-eng-review's self-check pointer escaped the rewrite (found by the setup lane's installed-pointer check). * Integrator: outside-voice fences use the shared runtime prelude; TMPDIR mktemp and per-round variant paths in design templates * Integrator: C8 router omits disabled skills on every render (relink renders the served Claude router in place); Windows-safe E4/E6 entries; investigate hooks under the cmd.exe check; browse_extension_id reader; order-insensitive test typecheck identities tsc printed one unchanged ceo-mode-option diagnostic's object members in a different order (content?/isError?) once lib/fs-utils joined the generator's import graph; identities now sort object-type members, and old baselines normalize on read. * Q1 (#2971 salvage): bound iOS proxy responses One wall-clock deadline covers headers and body (stalled and dribbling bodies time out), the whole body is capped at 64 MiB (upstream_response_too_large), every terminal path clears its timer, removes listeners and destroys the request, and the daemon threads the deadline through both listeners. Ambiguous mutations are still never replayed after a timeout or abort. Ported from #2971's proxy-transport work plus the size bound. * I1: grade the AUQ matrix's (recommended) marker on option labels, the field the auto-decide hook reads The decision-brief format puts a Pros / cons block with 'A) ... (recommended)' in the question text and the same marker on the option label. The matrix counted both and failed all six skills in periodic run 37151477069 with 'expected exactly one (recommended) option, found 2'. The question-preference hook reads labels only, so the gate now checks the native question: a Recommendation line in the question text and exactly one label ending in (recommended). Fixtures move to the native question shape and add the real spec capture plus a marker-only-in-text negative control. * Q1: type the proxy-transport test doubles (typecheck:test ratchet) * INV-1: gate outcome rows and troubleshooting sections for every lane's new messages * Release v1.91.19.0 (next free patch in the queue: #3022 holds 1.91.17.0, #3023 holds 1.91.18.0); migration renamed; engines.bun >=1.4.0 to match the CI pin (#2815) * v1.91.19.0 docs and metadata: CHANGELOG, TODOS, README troubleshooting/upgrade/type --selector, BROWSER.md type and profiles, AGENTS.md conventions, CONTRIBUTING #2949, touchfiles for the wave's new lib/bin files * I1: give hermetic Codex runs a temp dir beside CODEX_HOME so Codex can create its PATH aliases Codex 0.160 refuses to create its helper aliases (codex-linux-sandbox among them) when CODEX_HOME sits under its temp dir, and the runner put the hermetic home under the shard TMPDIR. Periodic run 37151477069 shows the warning and 'Unable to spawn codex-linux-sandbox ... No viable candidates found in PATH' for every exec_command in codex-sol-scope-termination. * I1: Codex evals fail a review whose commands never ran and stop reading tool errors as skill-load problems validateCodexReview accepted codex-review-findings in periodic run 37151477069 although its only command failed in bwrap ('I couldn't run the review'); it now requires execution evidence from classifyOutsideReview (B1's INV-1 classifier). codex-sol-scope-termination failed on 'apply_patch verification failed: invalid hunk' because the skill-load check matched any 'invalid'; it now matches Codex 0.160's skill-loader diagnostics only. * I1: run paid eval containers with seccomp unconfined so Codex's sandbox can start Docker's default seccomp profile blocks the user namespaces bubblewrap needs, so every Codex command in a paid eval failed with 'bwrap: No permissions to create a new namespace' (codex-review-findings, run 37151477069). Reproduced in the CI image locally: default profile fails, seccomp=unconfined runs the command. The runner stays non-root. * I1: shared-libs plan actor reads one exclusion grammar for scope lists Periodic run 37151477069 refused 'Existing-caller migration and helper hardening stay out of scope' as a hardening proposal: the actor had two partial exclusion rules (unanchored is/are/remains outside, and an anchored list that only stays unchanged), the third refusal at this boundary after 36633323521 and 36776104571. One anchored predicate now covers a bare list of scope nouns (hyphenated compounds included) with is/are/stays/remains and unchanged/untouched/excluded/out of scope/outside/not part; any other verb in the clause is still read as a proposal. Adds the real capture and two adversarial controls. * Parity caps: measured v1.91.19.0 union/skeleton sizes for ship, review, design-consultation, office-hours and design-shotgun, with the reason beside each * Regenerate generated output for v1.91.19.0 (gen:skill-docs --host all, goldens, llms.txt, context-budget fixture) * I1: plan-eng-review writes the Test Plan Artifact even while test choices are pending Periodic run 37151477069 (plan-eng-review-artifact): a non-interactive review recorded every Test review choice as pending and never wrote the QA test plan, because Step 5 said to write the artifact 'when these test and eval choices are resolved' while the artifact section says to list unresolved choices as pending. Both now say to write it after the decision gate, with unanswered choices under Pending Decisions. The pinned sentence becomes a meaning-level check. * I3: plan-eng-review states when findings first enter the report and that the scope menu's recommendation needs no repo probe Judge rationales for 'plan-eng-review/SKILL.md sections' (actionability 3.67 in periodic run 37151477069, clarity 3 in 36385945043 and 35567915613) cite two concrete tensions: 'do not add findings before Scope Challenge C' versus the ledger saves C itself performs, and 'Recommendation: A when a branch diff exists' versus 'do not probe for session state'. The write policy now says findings first enter through Scope Challenge C's ledger saves; the selector says its Recommendation line advises the user and needs no repository inspection. The pinned sentence becomes the new meaning-level regex. * I3: plan-ceo-review defines the closing sequence's cleanup step Two of three judge samples for 'plan-ceo-review/SKILL.md modes' in periodic run 37151477069 (actionability at the 4.0 gate) flag 'perform permitted cleanup' as never defined. The only cleanup the skill has is archiving old CEO plans, which needs approval; step 5 now says so. * I3: /qa-only's value-card rule names only its own card location qa-only rendered the /qa line 'Put it in the 8e.5 record (/qa) or under each proposed test (/qa-only)'; judge samples for 'qa-only/SKILL.md workflow' (actionability at the 4.0 gate in run 37151477069, 3.67 in 36907895813) flag the /qa-only-irrelevant 8e.5 record. The qa-mode bar now renders the invoking skill's location; both skills get shorter. * test-of-test ratchet: a repo script or setup read by path counts as product code codex-invocation-sites (runs scripts/gen-skill-docs.ts), gstack-config-gbrain-refresh (copies bin/<name> through a variable) and setup-cso-windows-probe (reads setup and scripts/build-cso-windows.ps1) exercise product code the detector could not see. * I1: DX floor actor strips the editor hint from the footer line only (from #2971) The custom-field reply normalization removed 'ctrl+g to edit in <editor>' from every line, so question text that resembled the menu footer could be rewritten and still authenticate. Only the last non-empty line is the footer. Ported from Garry's #2971 with its negative control. * I4: attribute every census testcase; deselections never count as SKIPPED The paid report counted every skipped JUnit testcase as SKIPPED and every testcase whose name was not a registry id as unattributed. On the 2026-10-03 census (run 37151477069) that was SKIPPED 82 / unattributed 72 (periodic) and SKIPPED 282 / unattributed 24 (gate), and every case-shard sibling also wrote a skipped history record that reset its pass-rate series. - attributeJUnitCase: a name is its registry id or CASE_TEST_NAMES label, else the only id its file registers, else the keyless file itself. Skipped other-tier cases, case-shard siblings, isolated or unselected cases and Bun's '(unnamed)' hook placeholders are zero-credit deselections with a reason and write no history record. Anything else is listed by name as unattributed. - One result and one history record per case per shard; a genuinely skipped selected case counts as SKIPPED and is listed with its reason, also in report-summary.md and collector-outcomes.json. SKIPPED panels show their reason and skipped trials are no longer called assertions. - CASE_TEST_NAMES labels for the three /cso cases. - tierSkipReason: a computed registration whose lane-tier ids are all registered to other test files is excluded at plan time, so the four gate and two periodic files that ran with every test skipped no longer take a runner. Census rerun: periodic SKIPPED 0 / unattributed 0, gate SKIPPED 0 / unattributed 0; verdicts unchanged. * I5: split the deploy file per case; guard PR-lane shard walls at 600 s test/skill-e2e-deploy.test.ts ran 668 s as one gate shard in the 2026-10-03 census (six sequential cases, 51-201 s each). It is now case-sharded, so each case is its own shard with its own (smaller) supervision budget; no timeout or budget was raised. - CASE_SHARDED_FILES gains the deploy file (names already equal ids). - Gate duration seeds for the six case keys from the census JUnit walls, so packing stays at 12 gate-census slices (CI cap 230m -> 200m). - shardDurationViolations + PR_LANE_SHARD_LIMIT_MS: a free test fails when a planned gate shard's recorded wall (committed seed, and the last gate census in test/fixtures/paid-report-census/gate-census-durations.json) exceeds 600 s. - Tests that pinned deploy as one shard follow the split (cookie planner keys, reuse fixture key, release/PR floors). * TODOS: /review fingerprint under Codex's read-only .git sandbox (found by the live Codex check) * Free-suite repairs from the first integrated Ubicloud run Product: - Outside-voice fences carry runtimeRootPrelude(ctx) explicitly again, so the resolver output runs on its own (the post-render pass skips fences that already assign GSTACK_ROOT); outsideVoiceRuntime stays deleted. - D5: an explicit CHROMIUM_PROFILE (gbrowser's per-workspace gbd) is never seeded from the shared profile or registered for `browse profiles prune`; this was also the free suite's write to the real ~/.gstack/browse-profiles.json. - gstack-ios-qa-daemon: BASH_COMPAT=50 for its >512-byte --help heredoc. Tests migrated to the wave's deliberate contract changes: - B1 heading admits `unverified`; sandbox flag is ${_GSTACK_CODEX_SANDBOX:?}; verdict-form output (VERDICT/FINDINGS/REASON); fake codex follows the preflight + stdin + -o form; Q2 notice marked shown in a private state root. - C1: an exported root needs bin/ and lib/; a lone GSTACK_BIN is not honored. - C4: plan-ceo-review is carved on every host (carrier, discovery and methodology fixtures read sections/ when the render has it). - E7: READY prints the resolved Aside path. - INV-3: the cookie-approval reconstruction reverses the bun-install mktemp template. - F1's `Plan:` binding read gets a reasoned tracker-guard exemption. - Browse stubs stop their server instead of scheduling process.exit; the headless launch check only counts browsers its own launch started. * Regenerate generated output after the outside-voice prelude repair * I1: revert the seccomp-only eval container option; it does not let Codex's sandbox start Run 37158847998 on this branch and a stock Ubicloud VM both show that seccomp=unconfined only moves bubblewrap's failure from 'No permissions to create a new namespace' to 'Failed to make / slave: Permission denied'. Codex's sandbox started only with the host sysctl kernel.apparmor_restrict_unprivileged_userns=0 plus seccomp and AppArmor unconfined on the container, which a container job cannot set from its steps. The relaxation bought nothing, so it goes; the fix is filed as a TODO for the integrator. This reverts commit ec97187. * I1: a Codex review eval needs a completed command, not just review-shaped text Run 37158847998 recorded codex-review-findings as passed again: the agent wrote 'I couldn't run the review because the execution sandbox failed (bwrap: Failed to make / slave)', the failure never reached stderr or a command event, and no phrase rule matched. validateCodexReview now also requires at least one command_execution event that completed with exit 0. Fixtures gain that event; the CI output is a new negative control. * Evals lane follow-ups: close the skillify gate TODO (I2 green), file its open reds, CHANGELOG lines; refresh measured parity annotations * Regenerate generated output after merging wave/evals * D2: bound the navigation guard's tab reset (commit, 5 s) Ubicloud run 3: a click to a link-local link reported the block, but the reset's page.goto('about:blank') waited the full 30 s default navigation timeout, so the command hung past the test deadline and the stale block leaked into the next command. The reset now waits for commit only, with a 5 s bound. * I1: run Codex paid slices on a host-run job where bubblewrap can start Docker's default seccomp/AppArmor profiles and the Ubuntu 24.04 host's kernel.apparmor_restrict_unprivileged_userns=1 keep Codex's bubblewrap sandbox from starting in a job container, so every Codex command failed. The planner now packs Codex shards into their own slices (plan.codexSlices, validated against the entries); evals-periodic runs those slices on a new host-run job that lifts the host restriction and runs the same CI image with seccomp and AppArmor unconfined. Artifact names match eval-slices, so the report reconciles one manifest. Gate plans have no Codex slices. * I1: the /codex review eval passes only when Codex executed Run 37158847998 recorded codex-review as passed while the sandbox preflight said unavailable and no review ran: the case only checked the session exit and an optional output file. It now requires the shared validator's executed verdict (findings, clean or unverified) in the codex command's own output and a written, non-fail-closed GATE. * TODOS: characterize the free-suite timing reds from Ubicloud run 4 * I1: QA evidence names the probes to rerun after an input change, before the one-shot materialize ship-exploratory-late-input returned inconclusive in gate census 37158847998: the fixture locale changed after captures 001 and 002, the agent reran only input 9, and materialize (which publishes once) recorded capture 001 as open. Its next line then said to rerun what is open, which the published verdict no longer allowed. Every capture now lists the commands that observed an older input snapshot and have no rerun on current inputs (revalidate, plus the next line), and materialize's next line says the verdict is final. * I1: the eng batching counter reads records under "Section N — <review> review" headings plan-eng-multi-finding-batching failed in 6 of 12 runs with reviewCount 0-1 (or a timeout with every ask counted as Step 0). In run 37158847998 the agent asked 11 separate questions, each with a saved ledger record, but filed records R2-R10 under '## Section 1 — Architecture review'. The counter strips 'Section 1' but not the dash separator, so it rejected every record outside the Decision ledger and the floor never filled. Quoted, historical and unrelated headings still do not count. * I1: plan-eng-review-artifact names its report file without making it the only output Runs 37151477069 and 37158847998 wrote the whole review, including the GSTACK REVIEW REPORT, into review-output.md and never ran the Test Plan Artifact block. #3017 made the artifact a hard assertion and dropped the prompt's explicit artifact instruction, leaving 'Write your review to review-output.md' as the one named destination. The prompt now names that file as the selected report file and says every other output keeps the destination the skill gives it, without naming the artifact. * browse: close returns when Chromium exits; Xvfb cleanup does not wait on its own zombie BrowserManager.close() detached every 'disconnected' listener before browser.close(). That also removed Playwright's internal listener that resolves close(), so every launched-mode close waited the full 5 s race timeout (measured 5,000 ms on origin/main and on this branch; raw Playwright closes in ~50 ms), and Playwright 1.62 has no Browser.process() for the SIGKILL fallback. The daemon's SIGINT shutdown therefore took over 5 s, which reddened the CLI lifecycle test under load (1 of 8 on main) and left browsers alive at shard cleanup. gstack now detaches only its own listeners: close takes ~50 ms. cleanupXvfb treated a terminated Xvfb child as alive until our own blocked event loop reaped it, so every close waited a full second then sent SIGKILL to a zombie. A zombie now counts as exited; the display test counts only running Xvfb processes. * I1: office-hours-brain-writeback reads the long skill with Read and skips the preamble 4 of 18 trials ended error_max_turns at the unchanged 12-turn budget with a gbrain search but no gbrain put. The failing trial in 37151477069 read the 1260-line rendered SKILL.md with cat, got truncated output, spent four more turns re-reading it with sed, then ran the preamble's skill-start block. The passing trials skipped the preamble. The prompt now says to read the skill with the Read tool and skip the preamble, onboarding, telemetry and contributor sections, as the plan-eng fixtures already do. Budget unchanged. * CI: no credential-shaped URLs in added lines (quality scan); release floor measured on the merged tree The quality job's credential scan of the PR diff flagged 7 db.url_with_password lines: three docs lines now describe the login instead of spelling a URL, and three tests build the URL from parts. paid-retry-supervision's periodic release floor is the merged tree's measured 35,711 (main's value; the evals lane's 33,821 predates main's new paid files). * qa-evidence: hold the stdout-pipe cases on a FIFO, not on a paused child stream Ubicloud run 4 finished the blocked-forever case in 71 ms (it should take the 5 s write deadline): the 2 MB write was not held, because saturation depended on how the runtime buffers a paused child pipe. On POSIX the child's stdout is now a FIFO the test reads only when it chooses to, so the kernel holds the write; Windows keeps the paused pipe. * TODOS/CHANGELOG: the three run-4 timing reds are fixed at their causes (daemon close latency, FIFO-held pipe tests) * I1: a QA helper call bun could not load is checked at the installed path, not failed outright review-exploratory-small-cli failed gate census 37162480720 with 'command outside declared caller observation interface': the actor dropped one character from the shard directory in a gstack-qa-evidence capture path, bun refused to load it ('Module not found', exit 1), and the next call ran the same capture correctly. A failed call whose only output is bun's refusal for a gstack-qa-evidence/deadline path is now held to the interface as the same command at the installed helper path, so a compound or undeclared command still fails. * I1: plan-eng-review's final self-check names its three outputs The fixture-prompt repair did not hold: in 37162480720 plan-eng-review-artifact again wrote one review-output.md, ran the skeleton's read-back gate on it and stopped without the QA Test Plan artifact (no gstack-paths/gstack-slug block). The transcript shows the agent's last check before finishing is the skeleton's Section self-check, which said only 'and outputs'. It now names the saved report, the QA Test Plan artifact and the Review Log, each at its own destination or presented as not persisted (the existing write policy). * I1: revert the plan-eng-review self-check wording; it did not change the artifact outcome A local run of plan-eng-review-artifact in the CI image (Claude Code 2.1.284) at e2f65ab22 still wrote one review-output.md in 8 turns and ran no Bash, so the QA Test Plan artifact block never ran. Keep the skill text unchanged; the case is deferred with its evidence. * I1: seed the Codex files' periodic durations from their first executing run Every Codex file recorded ~1 s because its commands could not start in the container (codex-e2e*, outside-voice, workflow). Run 37162480720 executed them on the host-run job; its slice 24 walls replace the placeholders so the planner packs the Codex slice by real work (7 files, ~4.9 min estimated). * Refresh scripts/free-test-durations.json (bun run test:ubicloud --record-durations; covers the wave's new test files) * evals2 follow-ups: outside-voice-evidence helper in five case touchfiles; close the evals lane's Codex-CI, late-input, batching and brain-writeback TODOs (run 37162480720); CHANGELOG lines * fix(browse): take the next free X display when a concurrent allocator wins the race Pick-then-spawn let two allocators (parallel daemons or test shards) choose the same free display; the losing Xvfb exited and the handoff failed (run 6, handoff 'failure before launch'). spawnFreeXvfb now treats a display taken during startup as a collision and moves on. Regression test forces the race with a delayed Xvfb wrapper (pre-fix: fails; two-process repro 13/15 -> 0/15). * test(ci): read the host-run Codex eval job's docker env, mounts and uploads eval-codex-slices runs on the host and passes its executor env through docker -e flags and HOME through a $RUNNER_TEMP/eval-home mount. The evidence and coordination tests now map those back to the container view instead of assuming job.container, and the job's native-capture upload carries the qa-callers paths like the other executors. * I1: plan-eng-review-artifact runs the real interactive review in a PTY, replacing the composing fixture The claude -p case told the model to skip the preamble and questions and go straight to the review. In 37151477069, 37158847998, 37162480720 and a local CI-image run it wrote the whole review into one file without executing the workflow, so it measured composing, not /plan-eng-review. The interactive batching PTY run in 37158847998 did write the artifact during Test review, so there is no product defect. New test/skill-e2e-plan-eng-artifact.test.ts keeps the case id, tier (periodic), kind (rule) and budget (CAPTURE_LONG_MS): it drives the real /plan-eng-review over the same dashboard plan and sources, answers each question with its first option, stops once the QA test plan exists or the report completes, and asserts the same requirement (exactly one new eng-review test plan in the run's own state root, about the dashboard change). The composing case is retired from skill-e2e-plan.test.ts in this commit. runPlanSkillCounting reports engTestPlans (read before fixture cleanup) and passes them to isCollectionComplete. One paid run in the CI image (Claude Code 2.1.284): pass, collection_complete at 498 s with one test plan. The new file takes its own periodic runner (seeded at 498 s), so the periodic census plans 26 slices: the planner's --max-parallel and both executors' caps move from 25 to 26 so every slice still starts at once; the paid-file census pin moves from 106 to 107. * evals2 final: plan-eng-review-artifact touchfiles point at its PTY file and helpers; plan file wall drops the moved case Applies the lane's touchfiles-data patch (its outside-voice-evidence hunks were already in 9a96fb3). skill-e2e-plan.test.ts lost one 600 s case to skill-e2e-plan-eng-artifact.test.ts, so its periodic wall is two 900 s + two 600 s (3,000,000 + reserve) and its timeout pin lists the remaining cases. * TODOS/CHANGELOG: close plan-eng-review-artifact (real PTY review, 498 s pass) * test: plan-eng-review-artifact is a PTY runner consumer (selected by claude-pty-runner and pty-screen changes) * test(windows): spawn cmd.exe hooks verbatim, use a .invalid UNC host, and match CRLF fixtures and native separators windows-free-tests reds on 364f0f3, each verified on windows-latest (46 pass, 1 POSIX-only skip, 0 fail): - investigate-freeze-path cmd.exe cases: Bun MSVC-quoted the command line cmd.exe parses, so cmd saw \"bash and failed; Node's shell spawn (what Claude Code uses) passes it verbatim. - investigate-freeze-path UNC cases: a single-label host made each filesystem probe wait on network name resolution (cd 11 s, -L 2.8 s); a reserved .invalid host fails at once (55-80 ms). - paid-report-census: the junit fixture is CRLF in a Windows checkout. - codex-invocation-sites: rendered paths use backslashes on Windows. * eval(plan-design-with-ui): the Step 0D focus title may spell the seven dimensions as a word Run 37166586458 slice 5 (dfaf154) asked 'D2 - Review all seven design dimensions, or focus on specific areas?' and reached its review finding, but the focus predicate only accepted the numeral, so the case failed with no acknowledged Design focus. Both recognizers accept 7 or seven; the byte-exact capture is a free regression, and engineering/other-count/ no-'design' titles stay rejected. * eval(ship-docsync): git check-ignore is a declared read in the docs observation interface Run 37166586458 slice 2 (dfaf154): ship-docsync-completion's parent ran 'git check-ignore -v .qa-state' to verify the prompt's claim that the observer probe is Git-excluded. The read was outside the closed Git list, so the run's DOC_PATH replacement was unprovable and failed. check-ignore cannot write; it joins the list and the interface prose. Free regression replays the captured command against the fixture and keeps composition and redirects rejected. * fix(browse): read Xvfb's own startup error to classify a lost display race Run 10 (1ffd38e) failed handoff's concurrent promotion once: 'Xvfb on :100 exited during startup (code 1)' with the display free again by the exit check, so a holder that won the lock and closed within the 100 ms poll looked like a broken install. Startup stderr now goes to a private file; 'Server is already active for display' is a collision and moves on, and any other startup exit carries Xvfb's fatal line. Regression: a stub that reports the collision for one display and exits leaves spawnFreeXvfb on the next display (pre-fix: the misleading install error). * test: merged-main fixtures (E6 copies the Claude render library gstack-config now sources; the census fixture's plan-mode-no-op budget follows its v1.91.18.0 policy) * qa: materialize runs once per PROBE_DIR, and qa-only says §1 reuses the REPORT_DIR it resolved Run 37170610789 slice 3 (feaa28d): the qa-only workflow judge scored clarity 3.33 and named two ambiguities: §4 said materialize runs once per report root while mixed runs materialize each surface's PROBE_DIR, and SKILL.md deferred exploratory §1 without saying §1 reuses the REPORT_DIR resolved under Prepare Report Artifacts. A local run of the same judge on the repaired text: clarity 3.67, completeness 4.33, actionability 4 (pass); all three samples call the workflow internally consistent. * Regenerate generated output for the qa materialize/REPORT_DIR wording * eval: three detector defects from run 37170610789 (feaa28d) - safety-codex-consult-embed: the fake codex recorded argv only, but every codex exec takes its prompt on stdin since E5; it records stdin for the '-' form and the check reads both. - ship-docsync-missing-asset: 'ls ... 2>&1' on a read-only command is the same harmless stderr wrapper as '2>/dev/null'; stdout redirects, pipes and script calls stay rejected. - qa-functional-webhook-fix: a pre-repair capture whose command reran on the repaired inputs may be classified superseded (exploratory §4); the evidence check accepts that and still rejects superseded on a post-repair capture or one never rerun. Each free regression fails without its repair. * fix(test:free): a browser record removed mid-read by a shutting-down daemon is absent Run 13 (feaa28d) needed the flaky-retry for 'owned detached browser settlement > cancel': 2 unattributed failures instead of 1. A 16-vCPU stress run reproduced it once in 64 with the cause: 'browser ownership unavailable'. On SIGINT the cancelled daemon removes its state file; record() saw it exist, then lstat/read hit ENOENT, which failure() reports as an ownership error. ENOENT there now means absent. New mode cancel-vanishing-record forces the removal between the check and the read (pre-fix: 2 unattributed, deterministic; now 1). * test(windows): the 2>&1 wrapper regression uses relative reads (no native paths in literal commands) * test(plan-seed): wait for the fixture CLI's slash event instead of a fixed 50 ms Run 14 (b2f729b) needed the flaky-retry for 'startup-placeholder': the events log had paste/enter/end_turn but not yet the slash 50 ms after it was sent. The check now polls (bounded 5 s); the deadline assertion measures when submission settled, which is what it bounds. One test typecheck diagnostic goes away with the typed reader. * eval(plan-design-with-ui): recognize the Step 0D focus question by meaning, not one title spelling The b2f729b gate run titled it "I've rated this plan 5/10 on design completeness. Review all 7 dimensions, or focus on specific areas?", the third native wording to miss a spelling-pinned regex. The native form is now a class: a title that offers all seven (numeral or word) dimensions or passes versus a focus and names design. Captured titles from 37166586458 and this run pass; other counts, engineering passes, titles without design or without a focus stay rejected. * eval(shared-libs-read-only): curl answers the provider rate-limit probe with the same 404 as gh Gate census 37174266054 slice 9: the model probed API availability with curl -o /dev/null https://api.github.com/rate_limit (GET, discarded). The gh transport already answers rate_limit as an unknown endpoint (404); the curl transport refused it, and the refusal counted as a read-only violation. curl now gets the same 404. POST, file output, other repos and raw hosts stay refused. Paid rerun (gate, 1 trial): PASS in 177 s. * eval(shared-libs-plan-callers): an exclusion may combine states (unchanged/out of scope) Periodic census 37174266054: every option ended 'Scheduler semantics, existing callers and helper hardening stay unchanged/out of scope', which the actor read as proposing hardening. Exclusion states may now be joined by '/', 'or' or 'and'; the captured question is answered, and clauses that pair a state with a real change ('stay unchanged/harden the helper') are still refused. Paid rerun (periodic, 1 trial): PASS in 296 s. * paid: --case plan-eng-review-artifact selects its PTY test by its registered name A local --case run matched 0 tests (no spend): the PTY test's Bun name is not its case id, and CASE_TEST_NAMES had no label for it. * eval(plan-eng-review-artifact): periodic checkpoint resumes at Test review; the fresh full run moves to marathon A fresh interactive /plan-eng-review cannot reach Test review inside the 600 s periodic budget: startup plus one-at-a-time Scope Challenge questions took 595 s with six questions and no Test review in census 37174266054 and in a local run (one earlier fresh run passed at 498 s). Per the eval-speed policy, the long run moves to the non-blocking marathon lane and the blocking case becomes a focused checkpoint. - plan-eng-review-artifact (periodic, 600 s unchanged): resumes the real interactive skill from a saved review record: the plan, the Scope Challenge result, Sections 1-2 complete, and a Decision ledger rebuilt from that census run's six actual native questions and answers (test/fixtures/plan-eng-artifact-resume-37174266054.json). The live skill runs Section 3, Test review. Assertion unchanged: exactly one new eng-review test plan in the run's own state root, about the dashboard change. Paid trial (local, Claude Code 2.1.284): PASS, 364 s. - plan-eng-review-artifact-full (marathon, PTY_LONG_MS): the fresh start-to-finish run with the same assertion; budget measured beside it. - test/plan-eng-resume.test.ts proves the rebuilt record is a valid resume point and rejects missing, unanswered, edited or duplicated records. - Registry: touchfiles, tiers, kinds, test-name labels, durations, paid file count 115, PTY consumer list; CHANGELOG and TODOS. * setup-deploy: say which deploy fields count as confirmed, how to resolve the default branch, and group the detection tests Periodic census 37176837432: the setup-deploy workflow judge's actionability mean was 3.67 (4, 4, 3). The low sample named the Fly/Heroku/Railway/GitHub Actions deploy trigger left undefined until Step 4's 'not already confirmed' catch-all and an unstated way to find <default branch>, and misread 'A || B && C' detection lines. Step 4 now defines a confirmed field and names the paths that leave the trigger unset, <default branch> resolves with gh repo view, and the two OR'd detection tests are grouped explicitly (same behavior). One judge run on the repaired text: 4/4/4, pass. * Regenerate generated output for the setup-deploy wording * TODOS: census data for the skipped CEO mode handoff (mode-routing, auto-decide-preserved); floor runner Edit-permission parse under collapsed TTY spaces * fix(browse): a blocked navigation that fails at once leaves the tab blank and the next command clean PR CI on 8d2d269 (free-suite 3): the redirect to [::ffff:169.254.170.2] failed immediately on the runner, so its chrome-error page committed around the guard's about:blank reset, and leaving the error page committed a second about:blank after the reset resolved, interrupting the next three tests' gotos. The guard now lets the blocked request settle (bounded at 1 s; a hanging target is still cut off by the reset), resets, and waits for the main frame to stay quiet before returning. Regression: a redirect to a blocked address on Chromium's unsafe port 1 fails at once (pre-fix: blank check and the next goto fail, deterministically). * qa: the checkpoint note is created for browser runs and composed for functional ones Periodic census 37178143007: the qa-only workflow judge scored actionability 3.67 again; one sample read step 2's 'Create exploration-NNN.json' as applying to functional runs before reaching 'Functional: do not write this file'. The sentence now names both surfaces up front; byte-neutral (the browser line drops its restated clause), qa union 1.1098 under its 1.110 cap. One judge run on the repaired text: 3.67/4/4, pass. * Regenerate generated output for the qa checkpoint wording * eval(section-loading): byte-range prints of a section count as a complete read Periodic census 37178143007: plan-ceo-section-loading printed review-sections.md in five byte ranges (head -c 16000, then tail -c +N | head -c 17000, one-byte overlaps) and the detector, which matched whole printed lines, saw the lines split at each range edge as missing. Outputs of the commands that name the section are also joined on their overlap. Regression: the captured ranges count; dropping a middle range or the last range does not (pre-fix: the complete ranges were refused). * test(plan-design-review): count design terms the review wrote into the plan * fix(aside): probe names the platform (GSTACK_PLATFORM override) so the macOS pitch never re-probes the OS * fix(plan-ceo-review): Step 0E mode handoff runs gstack-ceo-mode-handoff so the selected mode is printed, not only composed * chore: refresh context-budget fixture * test(auto-decide): accept Step 0E's documented AUTO_DECIDE line with its decisions suffix; replay the captured trial * docs: CHANGELOG and TODOS for the CEO mode handoff and platform line * test(plan-design-review): plan-term credit needs a rated summary; adapter regression for census 37179171083 * test(auto-decide): the printed handoff may precede Step 0E's provenance log; replay census 37182865432; TODOS for two rotating census reds * test(plan-ceo-section-loading): read the skill's ledger and currentDecision records for the stale-fill decision; replay three census reports * test: $B counts as a browser call only in command position; qa functional prompt defines its classification values * test(shared-libs-plan-callers): a verbless unchanged-scope list is an exclusion; replay census 37186854666; TODOS for two census reds * fix(settings-hook): retry a lock released between mkdir and the existence check instead of dropping the mutation * test(ship-docsync): the child's reply to a parent SendMessage counts as its spawned-marker report; replay PR run 37188668421 * test(browse): Xvfb stub-failure tests allocate from ranges no other allocator uses * perf(ship): documentation gate snapshots and compares its candidate with gstack-docs-candidate * fix(review): unverified Codex mode runs the passes in the mode summary too; stale-fill decision check accepts any ledger row ID * test(shared-libs-plan-callers): judge scope by forbidden verb+object actions, not an exclusion grammar; replay six refusals and four passing censuses * test(shared-libs-plan-callers): type the brief-line parameter * test(ship-docsync): an inline-code status quote is not a false current claim; replay PR run 37196811165 * fix(review,ship): Codex challenge and structured-review bodies run in unverified mode, matching their headers * docs(TODOS): overlay-harness control-arm error and format-approach provider refusals, with run links * docs(TODOS): qa-only workflow judge data point --------- Co-authored-by: Lewis <lewispeel@hotmail.com> Co-authored-by: Yugz <yannduzelier@gmail.com> Co-authored-by: Florian <florian.cleyetmerle@gmail.com> Co-authored-by: cdecook23 <decook23@gmail.com> Co-authored-by: mb1810 <michael.berger1810@gmail.com> Co-authored-by: Koda <koda@edgeviewfinance.com.au> Co-authored-by: Merit Blake <merit@newsprint.com> Co-authored-by: Nguyen Thanh Dat <ntdat812@gmail.com> Co-authored-by: pneumorea-ai <pneumorea@gmail.com> Co-authored-by: maxpetrusenkoagent <max.petrusenko.agent@gmail.com> Co-authored-by: Matt Van Horn <mvanhorn@gmail.com> Co-authored-by: Infiniteyieldai <infiniteyield.ai@gmail.com> Co-authored-by: Gabriel Rondon <grondon@gmail.com> Co-authored-by: Davon Johnson <d.johnson@dayofshow.net> Co-authored-by: Kevin Grasso <kevin@biblingo.com> Co-authored-by: Xuan-Ce Wang <261591073+xwang4-svg@users.noreply.github.com> Co-authored-by: loulanyue <260355617@qq.com> Co-authored-by: Thomas Reuser <thomasreuser@gmail.com> Co-authored-by: Bit0ps <alex.dreamscaper@gmail.com> Co-authored-by: oli548 <312288720+oli548@users.noreply.github.com> Co-authored-by: Slava <slava.istomin@tuta.io> Co-authored-by: ryanlaiwy <129729476+ryanlaiwy@users.noreply.github.com> Co-authored-by: 李翊楠 <yinanli1917@gmail.com> Co-authored-by: Mammad M. <mammad@digit.az> Co-authored-by: Lockyer <135391289+Lockyer228@users.noreply.github.com> Co-authored-by: Spooks444 <193448029+Spooks444@users.noreply.github.com> Co-authored-by: punksterlabs <77217235+punksterlabs@users.noreply.github.com> Co-authored-by: Jonathan Zhang <topcoder1@Mac-Studio.local> Co-authored-by: Marek Cermak <marek.cermak@strv.com> Co-authored-by: y$un_ <forrest.sun527@gmail.com> Co-authored-by: neallee <neallee@gmail.com> Co-authored-by: hardy <932821449@qq.com> Co-authored-by: afshaker <andrew@phone2.io> Co-authored-by: cube <frosimanuel@gmail.com> Co-authored-by: jacob-wang <jiayuw794@gmail.com> Co-authored-by: miao <miao@macmini-2.local> Co-authored-by: Jonah Berg <jonah.berg.g@gmail.com> Co-authored-by: Jayesh Betala <jayesh.betala7@gmail.com> Co-authored-by: Ryan Alberts <ryan.a.alberts@gmail.com> Co-authored-by: AlinValentin7 <AlinValentin7@users.noreply.github.com> Co-authored-by: Evgenii Lopatin <e75533@gmail.com> Co-authored-by: Jino <dgfrty00@gmail.com> Co-authored-by: snig-17 <santosh.m.tiwari@gmail.com> Co-authored-by: Sholto McNeilage <sholto@nodenarrative.com.au> Co-authored-by: Phil Hie <philhie@outlook.com> Co-authored-by: Sinabina <sinabina@Sinabinas-MacBook-Pro-4.local> Co-authored-by: Jimmy Zeisweiss <4261258+jzeisweiss@users.noreply.github.com> Co-authored-by: liutiming <timingliu@outlook.com> Co-authored-by: Brandon Pugsley <aibpaem@gmail.com> Co-authored-by: benjamin beres <benjamin.beres@bienpreter.com> Co-authored-by: Som Samantray <som.samantray@gmail.com> Co-authored-by: Vaughan <vaughan@cairnsitguy.com.au> Co-authored-by: Bharadwaj Mittapelli <bharadwajmittapelli@gmail.com>	2026-10-04 07:58:41 -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