Garry TanandClaude Fable 5
e76f65a8da
* fix: pin the claude CLI to an exact version in the CI image + tripwire The image installed @anthropic-ai/claude-code UNPINNED and rebuilt weekly 'to pick up CLI updates' — while bun sat carefully pinned at 1.3.13 two RUN lines above. The PTY harness screen-scrapes this CLI's TUI, and that drift broke it three separate times (welcome-screen wedge on 2.1.233, skillify HOME discovery on 2.1.237, guard/freeze hooks on 2.1.162), each debugged as a flake first. Pin 2.1.251 (current latest), bump deliberately via a PR that runs the PTY gate, and enforce with test/ci-image-cli-pin.test.ts: any global npm install in Dockerfile.ci without an exact @X.Y.Z pin fails the free suite. The weekly ci-image cron stays as a cheap tag self-heal. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: stamp the claude CLI version into every eval-store run record Three harness breakages were traced to claude-CLI TUI drift only after long flake hunts, because no run record said which CLI it actually exercised. EvalCollector now stamps claude_cli_version (claude --version, cached once per process, 'unknown' when the binary is absent) into both partial and finalized records — schema-additive optional field, no SCHEMA_VERSION bump. Correlating a flake wave with a CLI release becomes a grep over ~/.gstack/projects/<slug>/evals/ instead of archaeology. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: give the spinning-shard kill test load headroom (30s -> 90s) The test spawns and group-kills three real children (one a busy-loop burning a full core) while five sibling shard processes compete for eight vCPUs. Under full-suite load it blew bun's default 30s per-test ceiling at 30,009ms — while passing in isolation in 1.4s — and red the only required lane. Every assertion in it is event-based (statuses, group-kill proof, heartbeat lines); the sole latency claim is the <30s kill-deadline sanity bound, which stays. Explicit 90s headroom, not a weakened oracle. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: green-by-skip census — skip counts in the classifier, all-skipped labeling in the paid runner bun's 'Ran N tests' line COUNTS skipped tests, so a codex/gemini shard whose every test self-skipped (binary absent on the runner — true of every CI runner today) exits 0, dodges the hollow-shard guard, and reads as coverage in the weekly census. The classifier now parses bun's ' N skip' / ' N pass' recap lines; ShardOutcome carries skippedTests; formatSummary and the fail-closed slices report label an all-skipped pass explicitly: 'all N tests SKIPPED — verified nothing'. Status stays 'passed' (external service availability is host state, not a repo regression) but the census can no longer mistake absence for coverage. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * refactor: extract composite actions for eval-lane setup; surviving lanes gain the fail-fast registry verification 'Fix bun temp' x3, 'Restore deps' x5, 'Seed claude interactive config' x3, and 'Register gstack skills' x3 were byte-near-identical copies across the legacy matrix, the sliced lane, and the periodic lane — and only the MATRIX copy of register-skills carried the 19-line dangling-symlink + frontmatter fail-fast loop written after a silent 'Unknown command' + 35-min-timeout incident. Extract all four into .github/actions/ composites; the register composite carries the verification loop (generalized over the skill list), so the sliced and periodic lanes — the lanes that SURVIVE the matrix deletion — now inherit the check they had silently dropped. Matrix-job inline copies are left untouched: that job is deleted next. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: delete the legacy 17-row eval matrix — the sliced lane is the only paid lane Every PR paid twice: the hand-enumerated matrix (18 test files, 22.6 min, ~$21 API measured on run 33263204465) ran serialized AHEAD of the strictly superior sliced lane via 'needs: evals' — 35.5 min wall and ~2x paid spend for the same diff. 14 of 17 rows carried no tier:, so periodic Opus benchmarks leaked into every PR (the e2e-plan row alone: 12/12 tests, 21.7 min, $7.28 — the wall-clock bound of ALL of CI). Parity receipt (static, pre-deletion): the sliced lane's gate census (49 files, derived from the runner itself) strictly contains all 18 matrix test files, plus 31 files the matrix never ran. Pure deletion — one revert restores it. The PR comment moved into slices-report (same '## E2E Evals' upsert marker, now sourced from slice artifacts + carrying the fail-closed reconciliation verdict). plan-slices loses the needs edge; the dead workflow-level EVALS_TIER env goes with it. test/evals-workflow-matrix.test.ts (and its KNOWN_MATRIX_GAPS / KNOWN_TIER_UNSET burn-down ratchets — retired: the sliced census makes 'every gate file runs' true by construction) is rewritten as test/evals-workflow-wiring.test.ts: matrix stays deleted, planner/executor/ report tier + slice-count agreement, both surviving lanes on the shared register-skills composite with its fail-fast verification loop, PR comment survival. Expected: PR eval wall 35.5 -> ~13 min, per-PR paid spend ~halved. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: provider-runner timeouts kill the whole process GROUP; codex/gemini inherit the orphan-drain hardening All three provider runners (claude/codex/gemini) killed only the direct child on timeout: tool subprocesses the CLI spawned survived as orphans holding our pipes open and burning shared API rate (observed: a 600s timeout stretching past 1400s; a stalled run once burned a core for 15 hours). gstack-detach's watchdog had the same shape one level up — killpg SIGTERM, 5s grace, then a direct-child proc.kill() that orphaned grandchildren. Fix: spawn provider children via node:child_process with detached (own process group) and killProcessGroup(SIGKILL) in the timeout handler — runShardChild's proven pattern, EPERM/ESRCH fallbacks included. The codex and gemini copies also gain the reader.cancel() + stderr Promise.race hardening only the claude copy had (they still carried the blocked-drain hang it fixed). gstack-detach's watchdog now group-SIGKILLs after the grace. Regression net: test/session-runner-groupkill.test.ts drives the REAL runSkillTest against a fake claude shim (PATH override) that spawns a grandchild and wedges — the run must classify timeout within budget and leave neither shim nor grandchild alive — plus source pins on all three runners (detached + killProcessGroup, no bare timeout kill, no Bun.spawn reversion). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: skill-e2e-opus-47 renders SKILL.md fixtures into a mkdtemp — never the live tree mkEvalRoot ran gen-skill-docs with cwd=ROOT, regenerating every in-repo SKILL.md mid-run while concurrent paid shards copyFileSync those same files in their beforeAll (EVALS_JOBS>=4 locally, 2 per CI slice) — a sibling could capture a half-regenerated or opus-rendered SKILL.md, and a timeout before afterAll stranded the whole tree at the wrong model for every later shard. A cross-shard race that could flake ANY concurrent paid test. Render via the --out-dir flag gen-skill-docs grew for exactly this reason (mirrors the repo layout, which is all the fixture reads), read the skill heads from the render dir, delete it, and drop the afterAll restore-regen entirely. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: claude CLI version resolves in the runner parent, never on a test thread Eng-review finding: getClaudeCliVersion's fallback is a SYNCHRONOUS spawnSync on the same thread that polls concurrent PTY/session tests — the judgePtyState blocking class this overhaul kills elsewhere. The paid runner parent now resolves it once (cached) and stamps GSTACK_CLAUDE_CLI_VERSION into every shard's env; eval-store short-circuits on the env var, and the fallback spawn's budget tightens 10s -> 3s (bounded one-time stall, records 'unknown' on a slow CLI). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * test: wire skippedTests end-to-end through runPaidShard The census unit tests hand-built outcomes and the classifier tests parsed strings; nothing proved a real child's ' N skip' recap flows into outcome.skippedTests and the formatSummary label. A commandFor fake now prints the recap shape and the test asserts the parsed counts, the all-skipped predicate, and the 'verified nothing' label. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: make the setup composites rerun-safe (codex diff-review hardenings) restore-deps: 'cp -r SRC node_modules' with an existing node_modules NESTS the copy and leaves stale deps active — rm first. register-gstack-skills: 'ln -snf' hard-errors under set -eu when a REAL directory occupies the gstack slot — clear a non-symlink leftover first. CI workspaces are fresh today; a reusable composite must survive dirty reruns. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: sweep — every sync spawn in the test trees carries a timeout (436 sites, 157 files) spawnSync/execSync/Bun.spawnSync BLOCK the main thread, so bun's in-process per-test timeout can never fire while one waits — a hung child (stdin read, network probe, dead daemon) wedges the whole shard until the runner's external wall-clock SIGKILL. This exact class reached main: free-tests run 33262077256, test/gstack-memory-ingest.test.ts (normally 2.3s) held shard 2 at the 360s wall while its five siblings finished in ~65s. Mechanical sweep in two waves (12 + 4 fan-out agents, every edit verified against its call site): default timeout: 30_000 (matches the free runner's per-test budget), 120_000 for genuinely slow ops (installs, builds, playwright, provider CLIs), helper wrappers fixed ONCE where call sites route through them. Sites that only LOOK like calls (string fixtures, grep needles, comments) were skipped with reasons — the enforcement commit that follows marks them exempt. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * test: sync-spawn timeout tripwire — the wedge class stays extinct Free scanner over all test trees (test/, browse/test/, design/test/, make-pdf/test/, ios-qa, browser-skills): every spawnSync/execSync/ Bun.spawnSync call site must carry a timeout within a 30-line options window, or an explicit '// tripwire-exempt: <reason>' marker. Comment lines are skipped; exemptions are counted and ratcheted shrink-only (ceiling 6 = the 6 string-fixture/grep-needle sites where the pattern is CONTENT, not a call — marked in this commit). A scan-sanity test pins that the scanner still sees >100 real call sites so it can never rot to a vacuous green. Companion to the 436-site sweep in the previous commit. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: paid-lane flake telemetry — record-level attempts, flaky_retries, report surfacing bun --retry leaves a retried pass INVISIBLE in its output: a fail-then-pass prints the error detail but no (fail) result line and recaps as a clean pass (probed live on 1.3.10). So attempts are recorded where they cannot lie: EvalCollector.addTest stamps a 1-based attempt on same-name re-records (a retried test runs its body again and re-records), finalized runs carry flaky_retries, printSummary warns loudly, and the fail-closed slices report lists every passed-only-on-retry test — recorded and ranked, never blocking and never silent. Cross-model confirmed (codex reached the same don't-parse -the-stream conclusion independently). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: free-lane flake ledger — retry ON in CI, flaky-passes recorded and uploaded The runner's attribution-gated flaky-retry pass (cap 5, truncation veto) was OFF in the required lane and its FLAKY-PASS evidence was console-only — so a single timing flake red the merge gate while repeat offenders stayed unenumerable. free-tests.yml now sets GSTACK_FREE_RETRY_FLAKY=1 and points GSTACK_FLAKE_LEDGER at runner.temp; every flaky-pass appends a JSONL entry (SINGLE writer: the parent runner — no concurrent-append hazard by construction; fail-open with a loud warning so a broken ledger can never red the lane) and the artifact uploads UNCONDITIONALLY — a flaky-pass run is green, which is exactly when the evidence matters. Wiring pinned by free-tests-workflow-wiring; ledger behavior unit-tested incl. the fail-open path. Matches 2026 industry practice (retry for data, quarantine out of merge-blocking but never out of logging) with the repo's own receipts. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: eval:flake-rank — the flake-telemetry dial Aggregates per-test series across every finalized eval-store run (shard dirs included) plus the free flake ledger: runs, fails, RETRIED PASSES (the flake signature), avg duration — ranked retries-first. This is the readable dial behind two policies: a flaky pass never blocks a merge but is always ranked here, and the WS16 required-check promotion needs weeks of clean flake-rank, not vibes. --json for machines, --dir for downloaded CI artifacts. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: two-phase session timeout — silent APIs die at the startup grace, named The single spawn-armed timer charged API queue latency to the work budget: the recurring '0 turns / $0.00 / x3 attempts' failure with four budget-bump receipts (180->300s, 240->360s, 300->420s, 90->300s). Split: startup phase (no NDJSON byte yet) kills EARLY at min(grace, timeout) with the distinct exitReason 'timeout_startup' — an availability verdict, not transcript archaeology — and the work phase arms on the first byte for the REMAINING budget, so total wall never exceeds the timeout (tier envelopes are margin-free: tests pass timeout: CAPTURE_MS and bun-budget the same tier). Local grace 90s (observed queue latency 60-90s), CI floor 300s (TODOS-filed; shared runners queue harder), both pinned by the new grace tests with fake -claude shims covering the late-first-byte and silent-API paths. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: census integrity — 17 phantom selection keys deleted, reverse invariant added, gitignored dep patterns replaced, local map forks derived The merge-blocking gate census counted tests that could not run. Deleted (critic-verified against both quoted-occurrence and dep-registration liveness): 7 *-prosons-format keys with no declaring test, ship-plan- completion/-verification, review-plan-completion, design-shotgun-path/ session/full, autoplan-core (dead ~10 months), e2e-harness-audit (its namesake is a FREE-suite file), plus 2 dead LLM-judge keys and 2 free-file keys (budget-regression-pty, global-discover) misplaced in the PAID maps. Census: 191 -> 174 keys, gate 86 -> 78 honest. The new reverse invariant in touchfiles.test.ts makes the class structurally impossible: every key must be quoted in a living paid test file OR registered to an existing paid test file via its dep list (the constructed- name binding the 2026-08 self-registration sweep established) — zero exceptions needed today, with a live-file check on any future exception. Also: '.agents/skills/**' dep patterns replaced with the generator (scripts/gen-skill-docs.ts) — .agents/ is gitignored, so those patterns could NEVER match a git diff and review-template edits silently stopped selecting codex/gemini tests; the codex/gemini local touchfile maps are now DERIVED from the canonical map (loud throw if a key vanishes) instead of hand-forked copies that had already drifted. ios-qa-e2e demoted gate -> periodic: its gate declaration was never executable in CI (hardware exclusion only applies at tier=periodic), so every Linux PR planned a hollow shard. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: routing journeys lose their answer key and end at the routing decision The journey tests exist to catch skill-DESCRIPTION regressions (touchfiles: */SKILL.md.tmpl), but the fixture CLAUDE.md shipped an explicit prompt->skill lookup table — with the answer key in context, a badly regressed frontmatter description still routed correctly, so the tests could not fail on the exact class they select for. The fixture now carries only the generic invoke-skills nudge; the frontmatter carries the routing load. Also capped all 10 journeys at maxTurns 2 / tools [Skill, Read]: only the FIRST Skill call is asserted, so 5 turns of Read/Bash/Glob/Grep was pure spend — roughly halves each journey's cost. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: retire decided A/B experiments; vendor the pre-cut fixture; ban raw-SHA fixtures Three one-shot decision experiments kept re-running weekly as N=1 stochastic comparisons — flaky by construction with near-zero remaining information: skill-e2e-auq-repetition-cut-ab (its own header: gate "passed pre-landing, approved 2026-08-25"), skill-e2e-preamble-script-ab ("demoted post-Phase-3"), and opus-47's fanout arm-vs-arm (parA >= parB across two SINGLE stochastic runs — a coin flip). Deleted, with their selection keys; the SDK overlay-harness stays as the maintained instrument for the next experiment, and opus-47 keeps its routing-precision cases. verboseSkill() now reads the VENDORED test/fixtures/auq-pre-cut-...-SKILL.md instead of `git show ab66193e^:...` — a branch-local ref that dies on branch prune and already failed on shallow clones. New free tripwire (test/git-ref-fixture-tripwire.test.ts) bans the raw-SHA fixture class outright: quoted SHA:path rev-specs and gitRef-style hex defaults in the test trees fail the suite with the vendor-instead instruction. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: demote plan-ceo-review-expansion-energy to periodic Opus generator + a subjective 2-axis >=4/5 LLM-judge threshold sat in the MERGE-BLOCKING gate — the exact class its sibling posture tests were demoted for, with a receipt (a +21-line preamble change once flipped the score). CLAUDE.md's own tiering rule: Opus model test -> periodic. The weekly lane keeps the regression signal; merges stop paying a judge- temperament tax. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: paid shards get per-shard TMPDIR + CHROMIUM_PROFILE isolation and a kill-path cleanup backstop The free runner treats this isolation as MANDATORY (two concurrent shards on one Chromium profile kill each other's browser; shared tmp cross-contaminates) — the paid lane had none of it. Doubly load-bearing here: a shard that hits its 30-min wall is group-SIGKILLed, so per-test afterAll cleanup never runs; the rmSync backstop is the only thing keeping wedged runs from accumulating full git-repo workspaces in the shared tmpdir forever. This is the DAG prerequisite for raising EVALS_JOBS (next commit) — more concurrency on shared state amplifies exactly the shared-tree race class opus-47 exhibited. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat: paid-runner defaults 4x4 -> 8x2 — halve the local gate worst case 39 of 75 skill-e2e files hold exactly ONE test, so within-shard concurrency was dead weight for most shards: 4 jobs x 4 concurrency yielded only ~4-6 real in-flight sessions and a 13-wave local gate worst case (~6.5h). 8 jobs x 2 gives ~10-13 in-flight — under the documented-safe ~15 — and ~7 waves (~3.3h worst case). CI lanes keep their explicit EVALS_JOBS env (2 per slice; 4 for gate-census); this changes local defaults. Rollback trigger: sustained 429 storms in the WS1 telemetry across 2 PR cycles. test/eval-detach-timeout-floor.test.ts recomputed green (the raise LOWERS the worst-case floor). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: SHA-pin every action in the secrets-bearing eval lanes evals.yml and evals-periodic.yml execute PR-authored code with three provider API keys in env, yet rode mutable action tags (@v7/@v8/@v2/@v4) — while quality-gate.yml, osv-scanner.yml, and dependency-review.yml already model the SHA-pin pattern. All 30 uses sites across both lanes now pin the exact commit (tag noted in a trailing comment); dependabot's github-actions ecosystem keeps them fresh via PRs instead of silent tag moves. Pulled forward from the plan's endgame on the CEO-review + outside- voice agreement: supply-chain pins on secret lanes go first, not last. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: sweep wave 3 — the execFileSync family gets timeouts (90 sites, 17 files) The tripwire's regex covered spawnSync/execSync/Bun.spawnSync but not execFileSync — an entire blocking sync-spawn API family that could reintroduce the shard-wedge class undetected (ship review army). Same mechanical recipe as waves 1-2: timeout: 30_000 default, 120_000 for slow ops, shared wrappers fixed once, string-needle sites skipped with reasons. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * test: review-army + adversarial test hardening - Tripwire scans execFileSync too (ceiling 8: two more grep-needle string exemptions); merge-introduced timeout-less spawnSync in question-preference-hook fixed — the tripwire caught a site that landed on main AFTER the sweep, on its first day. - gstack-detach gains TWO watchdog kill regression tests: TERM-immune grandchild (the killpg-after-grace escalation) and the leader-dies variant (the pgid-at-spawn fix — the case the first test cannot see). - eval-flake-rank gets its unit suite (final-attempt accounting, artifact exclusion, shard recursion, recency bound). - Groupkill/startup-grace shim markers are per-run unique (pid-suffixed sleep durations): sibling Conductor worktrees run free suites with no machine lock, and fixed markers let one run pgrep/pkill the other's shims — a cross-run flake inside the anti-flake tests. - flake-ledger test pins the project-scoped local default; stale empty section headers in touchfiles-data deleted (they invited entries under deliberately retired categories). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: adversarial-review runtime fixes across the telemetry + kill paths - session-runner: exit-labeling keys off 'exit', not 'close' — an orphan holding the pipes could relabel a REAL exit (auth failure) as 'timeout_startup' availability noise; the kill path still always group-kills and cancels the reader (labeling and unblocking are separate concerns). Work phase arms on a flag, not firstResponseMs===0 (a same-ms first byte left the startup timer live all run). The CI startup grace is now a real FLOOR (Math.max), matching its name and pinning test. - gstack-detach: pgid captured AT SPAWN (== child pid under start_new_session) — resolving it after the grace raised ESRCH once the leader died on SIGTERM, orphaning TERM-immune grandchildren forever. - test-free-shards: ledger entries carry branch + git_sha (rev-parse split: '--abbrev-ref HEAD HEAD' printed the branch twice and recorded it as the sha); local ledger default is per-PROJECT, not the machine-global tmpdir. - eval-flake-rank: per-LINE ledger parse (one torn JSONL line vanished the whole series), 60-day recency bound (transcript-bearing files are MBs), shared isFinalizedEvalResultFile predicate (the artifact-taxonomy rule lived in three places); eval-store exports the predicate and finalize stops computing flakyRetries twice; paid-shards cleanup uses async rm (a SIGKILLed shard's git-workspace teardown blocked every sibling's stream classification on the parent event loop). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: CI trust-boundary + fail-closed repairs (adversarial findings) - Token/exec separation restored: slices-report (runs PR-authored code: bun install + the reconcile runner) drops to contents:read; the PR comment moves to a NEW slices-comment job holding the write token with ZERO repo code — no checkout, no bun, only downloaded artifacts + jq/gh. $GITHUB_ENV/BASH_ENV persistence is job-scoped, so the split is the boundary. The matrix-era report job had this property; the consolidation had regressed it. Pinned by the wiring test. - Reconcile exit captured via PIPESTATUS[0] in BOTH lanes: GitHub's default run-step shell has no pipefail, so `$?` after `| tee` was tee's exit — the fail-closed gate was silently fail-open. Wiring test pins it. - PR comment: final-attempt accounting restored the dropped COST accumulation (the dial read $0 forever), flaky passes render as the warning they are (never as failures), and a malformed tests[] artifact skips that file instead of aborting the whole comment under bash -e. - Remaining mutable action tags pinned (free-tests upload-artifact, ci-image checkout/docker trio — the image publisher holds packages:write and feeds the secret-bearing lanes). restore-deps fallback installs --frozen-lockfile; register-gstack-skills validates skill names before its rm -rf. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * chore: bump version and changelog (v1.77.0.0) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs: update project documentation for v1.77.0.0 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs: cross-model doc-review fixes — flake-ledger env knobs, CI retry-on note, stale version comment Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs: correct CHANGELOG receipt numbers to measured values Gate census keys: 78 -> 77 (bun-imported E2E_TIERS count). Sweep receipt: 586 sites/176 files -> 499 sites/146 files, measured by running this branch's spawnsync-timeout-tripwire against origin/main (exit 1, 499 violations across 146 unique files; green on this branch). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: slices-comment creates the PR comment via REST — the write-token job has no git context The token/exec split gives slices-comment NO checkout by design, and gh's pr-comment subcommand resolves the repo FROM git — it died with 'not a git repository' on PR #2746's first run (the update-existing PATCH path was already explicit-repo REST and worked). Create now posts through gh api repos/.../issues/N/comments, and the wiring test pins that no git-context-requiring comment call can creep back into the job. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: startup-grace probes clear CI for local semantics; new probe pins the floor clamp The two shim probes pass explicit 2s/4s graces, but in CI the runner clamps any explicit grace up to the 300s floor (deliberate adversarial-review fix), so 'silent API killed at the grace' died at the 30s work cap instead of 2s — a deterministic red on every CI run, green locally. The probes now pin LOCAL semantics with CI cleared (same save/restore pattern as their PATH shim), and a fourth probe pins the clamp itself: CI=1 + 2s grace + 6s timeout must kill at the 6s cap, still in the startup phase — proof an explicit low grace cannot bypass the floor. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
gstack
"I don't think I've typed like a line of code probably since December, basically, which is an extremely large change." — Andrej Karpathy, No Priors podcast, March 2026
When I heard Karpathy say this, I wanted to find out how. How does one person ship like a team of twenty? Peter Steinberger built OpenClaw — 247K GitHub stars — essentially solo with AI agents. The revolution is here. A single builder with the right tooling can move faster than a traditional team.
I'm Garry Tan, President & CEO of Y Combinator. I've worked with thousands of startups — Coinbase, Instacart, Rippling — when they were one or two people in a garage. Before YC, I was one of the first eng/PM/designers at Palantir, cofounded Posterous (sold to Twitter), and built Bookface, YC's internal social network.
gstack is my answer. I've been building products for twenty years, and right now I'm shipping more products than I ever have. In the last 60 days: 3 production services, 40+ shipped features, part-time, while running YC full-time. On logical code change — not raw LOC, which AI inflates — my 2026 run rate is ~810× my 2013 pace (11,417 vs 14 logical lines/day). Year-to-date (through April 18), 2026 has already produced 240× the entire 2013 year. Measured across 40 public + private garrytan/* repos including Bookface, after excluding one demo repo. AI wrote most of it. The point isn't who typed it, it's what shipped.
The LOC critics aren't wrong that raw line counts inflate with AI. They are wrong that normalized-for-inflation, I'm less productive. I'm more productive, by a lot. Full methodology, caveats, and reproduction script: On the LOC Controversy.
2026 — 1,237 contributions and counting:
2013 — when I built Bookface at YC (772 contributions):
Same person. Different era. The difference is the tooling.
gstack is how I do it. It turns Claude Code into a virtual engineering team — a CEO who rethinks the product, an eng manager who locks architecture, a designer who catches AI slop, a reviewer who finds production bugs, a QA lead who opens a real browser, a security officer who runs OWASP + STRIDE audits, and a release engineer who ships the PR. Twenty-three specialists and eight power tools, all slash commands, all Markdown, all free, MIT license.
This is my open source software factory. I use it every day. I'm sharing it because these tools should be available to everyone.
Fork it. Improve it. Make it yours. And if you want to hate on free open source software — you're welcome to, but I'd rather you just try it first.
Who this is for:
- Founders and CEOs — especially technical ones who still want to ship
- First-time Claude Code users — structured roles instead of a blank prompt
- Tech leads and staff engineers — rigorous review, QA, and release automation on every PR
Quick start
- Install gstack (30 seconds — see below)
- Run
/office-hours— describe what you're building - Run
/plan-ceo-reviewon any feature idea - Run
/reviewon any branch with changes - Run
/qaon your staging URL - Stop there. You'll know if this is for you.
Install — 30 seconds
Requirements: Claude Code, Git, Bun v1.0+, Node.js (Windows only)
Step 1: Install on your machine
Open Claude Code and paste this. Claude does the rest.
Install gstack: run
git clone --single-branch --depth 1 https://github.com/garrytan/gstack.git ~/.claude/skills/gstack && cd ~/.claude/skills/gstack && ./setupthen add a "gstack" section to CLAUDE.md that says to use the /browse skill from gstack for all web browsing, never use mcp__claude-in-chrome__* tools, and lists the available skills: /office-hours, /plan-ceo-review, /plan-eng-review, /plan-design-review, /design-consultation, /design-shotgun, /design-html, /review, /ship, /land-and-deploy, /canary, /benchmark, /browse, /connect-chrome, /qa, /qa-only, /design-review, /setup-browser-cookies, /setup-deploy, /setup-gbrain, /retro, /investigate, /document-release, /document-generate, /codex, /cso, /autoplan, /plan-devex-review, /devex-review, /careful, /freeze, /guard, /unfreeze, /gstack-upgrade, /learn. Then ask the user if they also want to add gstack to the current project so teammates get it.
Step 2: Team mode — auto-update for shared repos (recommended)
From inside your repo, paste this. Switches you to team mode, bootstraps the repo so teammates get gstack automatically, and commits the change:
(cd ~/.claude/skills/gstack && ./setup --team) && ~/.claude/skills/gstack/bin/gstack-team-init required && git add .claude/ CLAUDE.md && git commit -m "require gstack for AI-assisted work"
No vendored files in your repo, no version drift, no manual upgrades. Every Claude Code session starts with a fast auto-update check (throttled to once/hour, network-failure-safe, completely silent).
Swap required for optional if you'd rather nudge teammates than block them.
OpenClaw
OpenClaw spawns Claude Code sessions via ACP, so every gstack skill just works when Claude Code has gstack installed. Paste this to your OpenClaw agent:
Install gstack: run
git clone --single-branch --depth 1 https://github.com/garrytan/gstack.git ~/.claude/skills/gstack && cd ~/.claude/skills/gstack && ./setupto install gstack for Claude Code. Then add a "Coding Tasks" section to AGENTS.md that says: when spawning Claude Code sessions for coding work, tell the session to use gstack skills. Include these examples — security audit: "Load gstack. Run /cso", code review: "Load gstack. Run /review", QA test a URL: "Load gstack. Run /qa https://...", build a feature end-to-end: "Load gstack. Run /autoplan, implement the plan, then run /ship", plan before building: "Load gstack. Run /office-hours then /autoplan. Save the plan, don't implement."
After setup, just talk to your OpenClaw agent naturally:
| You say | What happens |
|---|---|
| "Fix the typo in README" | Simple — Claude Code session, no gstack needed |
| "Run a security audit on this repo" | Spawns Claude Code with Run /cso |
| "Build me a notifications feature" | Spawns Claude Code with /autoplan → implement → /ship |
| "Help me plan the v2 API redesign" | Spawns Claude Code with /office-hours → /autoplan, saves plan |
See docs/OPENCLAW.md for advanced dispatch routing and the gstack-lite/gstack-full prompt templates.
Native OpenClaw Skills (via ClawHub)
Four methodology skills that work directly in your OpenClaw agent, no Claude Code session needed. Install from ClawHub:
clawhub install gstack-openclaw-office-hours gstack-openclaw-ceo-review gstack-openclaw-investigate gstack-openclaw-retro
| Skill | What it does |
|---|---|
gstack-openclaw-office-hours |
Product interrogation with 6 forcing questions |
gstack-openclaw-ceo-review |
Strategic challenge with 4 scope modes |
gstack-openclaw-investigate |
Root cause debugging methodology |
gstack-openclaw-retro |
Weekly engineering retrospective |
These are conversational skills. Your OpenClaw agent runs them directly via chat.
Other AI Agents
gstack works on 10 AI coding agents, not just Claude. Setup auto-detects which agents you have installed:
git clone --single-branch --depth 1 https://github.com/garrytan/gstack.git ~/gstack
cd ~/gstack && ./setup
Or target a specific agent with ./setup --host <name>:
| Agent | Flag | What you get |
|---|---|---|
| OpenAI Codex CLI | --host codex |
Full install → ${CODEX_HOME:-~/.codex}/skills/gstack-*/ |
| OpenCode | --host opencode |
Full install → ~/.config/opencode/skills/gstack-*/ |
| Cursor | --host cursor |
Full install → ~/.cursor/skills/gstack-*/ |
| Factory Droid | --host factory |
Full install → ~/.factory/skills/gstack-*/ |
| Kiro | --host kiro |
Full install → ~/.kiro/skills/gstack-*/ |
| Slate | --host slate |
Pointer to the Claude install (Slate reads .claude/skills as a fallback) |
| OpenClaw | --host openclaw |
ACP spawn pointers + methodology artifacts via gen:skill-docs --host openclaw + the instruction-only digest below (full guide: docs/OPENCLAW.md) |
| Hermes | --host hermes |
Methodology artifacts via gen:skill-docs --host hermes + the instruction-only digest below |
| GBrain (mod) | --host gbrain |
Brain-aware skill variants, shipped from the GBrain repo |
Instruction-only tier (any rules-reading agent — Zed, Amp, Jules, side projects):
copy the 2KB digest at agents-digest/gstack-AGENTS.md
into a location your agent reads (for example, append it to your project's AGENTS.md).
It carries gstack's ethos, reuse ladder, and voice rules — no install required. The
digest's first line shows its gstack version; re-copy it after upgrading.
For Codex, setup reads the top-level model from
${CODEX_HOME:-~/.codex}/config.toml and generates the matching behavioral
profile. gpt-5.6-sol automatically receives bounded-scope instructions that
finish the requested lake without expanding into adjacent cleanup or speculative
hardening. The Sol profile is exact-match only: dated snapshots and other 5.6
variants get the generic GPT profile, and setup warns on near-misses like
gpt-5.6-sol-2026-08-01. Override detection with ./setup --host codex --model <id> — the
override applies to that run only; set model in your Codex config.toml to
make it stick across upgrades. After changing your Codex model, rerun
./setup --host codex to regenerate the skills.
Want to add support for another agent? See docs/ADDING_A_HOST.md. It's one TypeScript config file, zero code changes.
See it work
You: I want to build a daily briefing app for my calendar.
You: /office-hours
Claude: [asks about the pain — specific examples, not hypotheticals]
You: Multiple Google calendars, events with stale info, wrong locations.
Prep takes forever and the results aren't good enough...
Claude: I'm going to push back on the framing. You said "daily briefing
app." But what you actually described is a personal chief of
staff AI.
[extracts 5 capabilities you didn't realize you were describing]
[challenges 4 premises — you agree, disagree, or adjust]
[generates 3 implementation approaches with effort estimates]
RECOMMENDATION: Ship the narrowest wedge tomorrow, learn from
real usage. The full vision is a 3-month project — start with
the daily briefing that actually works.
[writes design doc → feeds into downstream skills automatically]
You: /plan-ceo-review
[reads the design doc, challenges scope, runs 10-section review]
You: /plan-eng-review
[ASCII diagrams for data flow, state machines, error paths]
[test matrix, failure modes, security concerns]
You: Approve plan. Exit plan mode.
[writes 2,400 lines across 11 files. ~8 minutes.]
You: /review
[AUTO-FIXED] 2 issues. [ASK] Race condition → you approve fix.
You: /qa https://staging.myapp.com
[opens real browser, clicks through flows, finds and fixes a bug]
You: /ship
Tests: 42 → 51 (+9 new). PR: github.com/you/app/pull/42
You said "daily briefing app." The agent said "you're building a chief of staff AI" — because it listened to your pain, not your feature request. Eight commands, end to end. That is not a copilot. That is a team.
The sprint
gstack is a process, not a collection of tools. The skills run in the order a sprint runs:
Think → Plan → Build → Review → Test → Ship → Reflect
Each skill feeds into the next. /office-hours writes a design doc that /plan-ceo-review reads. /plan-eng-review writes a test plan that /qa picks up. /review catches bugs that /ship verifies are fixed. Nothing falls through the cracks because every step knows what came before it.
| Skill | Your specialist | What they do |
|---|---|---|
/office-hours |
YC Office Hours | Start here. Six forcing questions that reframe your product before you write code. Pushes back on your framing, challenges premises, generates implementation alternatives. Design doc feeds into every downstream skill. |
/plan-ceo-review |
CEO / Founder | Rethink the problem. Find the 10-star product hiding inside the request. Four modes: Expansion, Selective Expansion, Hold Scope, Reduction. |
/plan-eng-review |
Eng Manager | Lock in architecture, data flow, diagrams, edge cases, and tests. Forces hidden assumptions into the open. |
/plan-design-review |
Senior Designer | Rates each design dimension 0-10, explains what a 10 looks like, then edits the plan to get there. AI Slop detection. Interactive — one AskUserQuestion per design choice. |
/plan-devex-review |
Developer Experience Lead | Interactive DX review: explores developer personas, benchmarks against competitors' TTHW, designs your magical moment, traces friction points step by step. Three modes: DX EXPANSION, DX POLISH, DX TRIAGE. 20-45 forcing questions. |
/design-consultation |
Design Partner | Build a complete design system from scratch. Researches the landscape, proposes creative risks, generates realistic product mockups. |
/review |
Staff Engineer | Find the bugs that pass CI but blow up in production. Auto-fixes the obvious ones. Flags completeness gaps. Advisory simplification lens flags over-built code — never blocks, never auto-applies. |
/investigate |
Debugger | Systematic root-cause debugging. Iron Law: no fixes without investigation. Traces data flow, tests hypotheses, stops after 3 failed fixes. |
/design-review |
Designer Who Codes | Same audit as /plan-design-review, then fixes what it finds. Atomic commits, before/after screenshots. |
/devex-review |
DX Tester | Live developer experience audit. Actually tests your onboarding: navigates docs, tries the getting started flow, times TTHW, screenshots errors. Compares against /plan-devex-review scores — the boomerang that shows if your plan matched reality. |
/design-shotgun |
Design Explorer | "Show me options." Generates 4-6 AI mockup variants, opens a comparison board in your browser, collects your feedback, and iterates. Taste memory learns what you like. Repeat until you love something, then hand it to /design-html. |
/design-html |
Design Engineer | Turn a mockup into production HTML that actually works. Pretext computed layout: text reflows, heights adjust, layouts are dynamic. 30KB, zero deps. Detects React/Svelte/Vue. Smart API routing per design type (landing page vs dashboard vs form). The output is shippable, not a demo. |
/qa |
QA Lead | Test your app, find bugs, fix them with atomic commits, re-verify. Auto-generates regression tests for every fix. |
/qa-only |
QA Reporter | Same methodology as /qa but report only. Pure bug report without code changes. |
/pair-agent |
Multi-Agent Coordinator | Share your browser with any AI agent. One command, one paste, connected. Works with OpenClaw, Hermes, Codex, Cursor, or anything that can curl. Each agent gets its own tab. Auto-launches headed mode so you watch everything. Auto-starts ngrok tunnel for remote agents. Scoped tokens, tab isolation, rate limiting, activity attribution. |
/cso |
Chief Security Officer | OWASP Top 10 + STRIDE threat model. Zero-noise: 17 false positive exclusions, 8/10+ confidence gate, independent finding verification. Each finding includes a concrete exploit scenario. |
/ship |
Release Engineer | Sync main, run tests, audit coverage, push, open PR. Bootstraps test frameworks if you don't have one. |
/land-and-deploy |
Release Engineer | Merge the PR, wait for CI and deploy, verify production health. One command from "approved" to "verified in production." |
/canary |
SRE | Post-deploy monitoring loop. Watches for console errors, performance regressions, and page failures. |
/benchmark |
Performance Engineer | Baseline page load times, Core Web Vitals, and resource sizes. Compare before/after on every PR. |
/document-release |
Technical Writer | Update all project docs to match what you just shipped. Catches stale READMEs automatically. Builds a Diataxis coverage map (reference / how-to / tutorial / explanation) so gaps are visible in the PR body. |
/document-generate |
Documentation Author | Generate missing docs from scratch using the Diataxis framework. Researches the codebase first, then writes reference / how-to / tutorial / explanation docs that actually match the code. Invokable standalone or chained from /document-release when the coverage map finds gaps. Learn more: tutorial • how-to • why Diataxis. |
/retro |
Eng Manager | Team-aware weekly retro. Per-person breakdowns, shipping streaks, test health trends, growth opportunities. /retro global runs across all your projects and AI tools (Claude Code, Codex, Gemini). |
/browse |
QA Engineer | Give the agent eyes. Real Chromium browser, real clicks, real screenshots. ~100ms per command. /open-gstack-browser launches GStack Browser with sidebar, anti-bot stealth, and auto model routing. |
/setup-browser-cookies |
Session Manager | Import cookies from your real browser (Chrome, Arc, Brave, Edge) into the headless session. Test authenticated pages. |
/autoplan |
Review Pipeline | One command, fully reviewed plan. Runs CEO → design → DX → eng review automatically (eng always last, so the shipping gate reviews the final amended plan) with encoded decision principles. Surfaces only taste decisions for your approval. |
/spec |
Spec Author | Turn vague intent into a precise, executable spec in five phases (why, scope, technical with mandatory code-reading, draft, file). Codex quality gate before file (blocks below 7/10), fail-closed secret redaction, dedupe against existing issues, archive to $GSTACK_STATE_ROOT/projects/$SLUG/specs/ for team-corpus recall. --execute spawns claude -p in a fresh worktree; /ship auto-closes the source issue on merge. Plan-mode aware. |
/learn |
Memory | Manage what gstack learned across sessions. Review, search, prune, and export project-specific patterns, pitfalls, and preferences. Learnings compound across sessions so gstack gets smarter on your codebase over time. |
/make-pdf |
Publisher | Markdown in, publication-quality document out. Mermaid and excalidraw fences render as vector diagrams, fully offline. Images scale to the page and never truncate; wide diagrams get their own landscape page. --to html emits one self-contained file, --to docx a Word doc. |
/diagram |
Diagram Maker | English in, editable diagram out. Emits a triplet: mermaid source, .excalidraw you can open and edit on excalidraw.com (hand-drawn style), and rendered SVG/PNG. Zero network. Embed the source in markdown and /make-pdf renders it. |
Which review should I use?
| Building for... | Plan stage (before code) | Live audit (after shipping) |
|---|---|---|
| End users (UI, web app, mobile) | /plan-design-review |
/design-review |
| Developers (API, CLI, SDK, docs) | /plan-devex-review |
/devex-review |
| Architecture (data flow, perf, tests) | /plan-eng-review |
/review |
| All of the above | /autoplan (runs CEO → design → DX → eng, auto-detects which apply; eng always last) |
— |
Power tools
| Skill | What it does |
|---|---|
/codex |
Second Opinion — independent code review from OpenAI Codex CLI. Three modes: review (pass/fail gate), adversarial challenge, and open consultation. Cross-model analysis when both /review and /codex have run. |
/careful |
Safety Guardrails — warns before destructive commands (rm -rf, DROP TABLE, force-push). Say "be careful" to activate. Override any MEDIUM warning; root/home recursive deletes and default-branch force-pushes are hard-denied. |
/freeze |
Edit Lock — restrict file edits to one directory. Prevents accidental changes outside scope while debugging. |
/guard |
Full Safety — /careful + /freeze in one command. Maximum safety for prod work. |
/unfreeze |
Unlock — remove the /freeze boundary. |
/open-gstack-browser |
GStack Browser — launch GStack Browser with sidebar, anti-bot stealth, auto model routing (Sonnet for actions, Opus for analysis), one-click cookie import, and Claude Code integration. Clean up pages, take smart screenshots, edit CSS, and pass info back to your terminal. |
/setup-deploy |
Deploy Configurator — one-time setup for /land-and-deploy. Detects your platform, production URL, and deploy commands. |
/setup-gbrain |
GBrain Onboarding — from zero to running gbrain in under 5 minutes. PGLite local, Supabase existing URL, or auto-provision a new Supabase project via Management API. MCP registration for Claude Code + per-repo trust triad (read-write/read-only/deny). Full guide. |
/sync-gbrain |
Keep Brain Current — re-index this repo's code into gbrain via gbrain sources add + gbrain sync --strategy code, refresh the ## GBrain Search Guidance block in CLAUDE.md, and auto-remove guidance when the capability check fails. --incremental (default), --full, --dry-run. Idempotent; safe to re-run. |
/gstack-upgrade |
Self-Updater — upgrade gstack to latest. Detects global vs vendored install, syncs both, shows what changed. |
/ios-qa |
iOS Live-Device QA (v1.43.0.0+) — drive a real iPhone over USB CoreDevice via an embedded StateServer in the app. Read Swift source, codegen typed @Observable accessors, run the agent loop. Optional --tailnet flag exposes the device to OpenClaw or any HTTP-capable agent on your Tailscale tailnet so remote agents can run iOS QA without ever touching the hardware. Capability-tier allowlist (observe/interact/mutate/restore), per-device session lock, audit log. |
/ios-fix, /ios-design-review, /ios-clean, /ios-sync |
iOS bug-fix loop, designer's-eye HIG audit, debug-bridge cleanup, and accessor resync. See docs/skills.md. End-to-end walkthrough: docs/howto-ios-testing-with-gstack.md. |
Standalone binaries
Beyond the slash-command skills, gstack ships standalone CLIs for workflows that don't belong inside a session:
| Command | What it does |
|---|---|
gstack-model-benchmark |
Cross-model benchmark — run the same prompt through Claude, GPT (via Codex CLI), and Gemini; compare latency, tokens, cost, and (optionally) LLM-judge quality score. Auth detected per provider, unavailable providers skip cleanly. Output as table, JSON, or markdown. --dry-run validates flags + auth without spending API calls. |
gstack-taste-update |
Design taste learning — writes approvals and rejections from /design-shotgun into a persistent per-project taste profile. Decays 5%/week. Feeds back into future variant generation so the system learns what you actually pick. |
gstack-egress |
Egress receipt auditor — every gstack-initiated off-machine send writes a tamper-evident, hash-chained receipt to ~/.gstack/security/egress.jsonl before the send. list shows what gstack attempted to send and to which host, grants shows the standing consent settings plus the exact command that revokes each, verify recomputes the hash chain and exits 3 on tamper (catches edits, reordering, and mid-chain deletion; truncating or deleting the ledger itself is out of scope — it's a forensic log, not tamper-proof storage). |
gstack-context-bill |
Token bill-of-materials — read-only, offline audit of what an installed skills tree costs in tokens: always-on frontmatter every session pays vs per-invocation SKILL.md + forced references. --diff compares two trees, --budget enforces a ceiling, --exact opts into Anthropic count_tokens (sends file text off-machine; writes an egress receipt first, degrades to the offline estimate if the receipt can't be written). |
gstack-code-intelligence |
Code-intelligence provider picker — wraps GBrain, Sourcebot, and Graphify behind one interface: options/status to see what's available, select to pick one, index/search to use it, suggest to check whether the one-time indexing offer should fire here. The offer triggers on large repos (1,000+ tracked files; a decline is persisted). Non-local providers refuse to index or search until you record per-repo consent (consent <repo> yes|no — the query text is repo-derived content), the per-repo trust policy's deny and read-only tiers veto write-class operations regardless of consent, and every off-machine send writes an egress receipt. Fully optional — with nothing selected, gstack falls back to grep. |
gstack-verify-gate |
Verification stop hook (opt-in) — blocks a Claude Code turn from ending until the project's declared verify command passes (after 3 blocked re-entries it yields with a loud still-RED warning instead of looping forever). Declare it on one line in CLAUDE.md: <!-- gstack:verify: bun test -->. Hooks bypass the permission system, so a declared command never runs until you trust it once per repo (gstack-verify-gate --trust); editing the command invalidates trust until re-granted, and every grant is audit-logged. ./setup never registers it for you — opt in with gstack-settings-hook add-event --event Stop --command ~/.claude/skills/gstack/bin/gstack-verify-gate --source verify-gate, remove with gstack-settings-hook remove-source --source verify-gate. |
gstack-wtree |
Working-tree fingerprint — prints a content hash of what's actually on disk (temp index seeded from the stat cache, ~40x cheaper than a full re-hash; untracked source counts, gitignored scratch doesn't). Identical content fingerprints identically through commits, rebases, amends, and squashes — it's what binds reviews and test evidence to content instead of commit SHAs. |
gstack-evidence |
Verification-evidence ledger — run --label <lane> -- <cmd> transparently wraps any test command (the child's exit code always passes through) and records what ran against which working-tree fingerprint; check grades each label FRESH/STALE/MISSING with --expect-cmd, --max-age, and --allow-paths binding. /ship and /land-and-deploy cite fresh evidence instead of re-running suites. Per-run logs are 0600, capped at 2MB, pruned after 30 days; the ledger and logs stay machine-local by design. |
gstack-issue-guard |
Tracker-text trust envelope — fetches GitHub issue/PR text (issue <n>, pr-body, pr-comments, or --stdin) and wraps it in a labeled envelope so agents treat it as data: injection-shaped lines get labeled even through fullwidth and invisible-character evasion, and forged envelope banners are defused. Every tracker-text ingress in gstack routes through it, enforced by a CI scanner. |
gstack-ios-qa-daemon |
iOS QA daemon — Mac-side broker between an agent and a connected iPhone over USB CoreDevice. Loopback by default; --tailnet opens a Tailscale-facing listener with identity-gated capability tiers. Single-instance via flock on ~/.gstack/ios-qa-daemon.pid. See docs/howto-ios-testing-with-gstack.md. |
gstack-ios-qa-mint |
iOS allowlist manager — owner-grant CLI for the tailnet allowlist. grant/revoke/list against ~/.gstack/ios-qa-allowlist.json (mode 0600). Remote agents never auto-allowlist; this is the explicit-intent path. |
gstack-ios-qa-regen |
iOS bridge regenerator — deterministically installs the canonical DebugBridge package, generates typed state accessors, and records the installed gstack version. Safe to rerun after source changes or upgrades. |
./setup also registers one default-on Stop hook in ~/.claude/settings.json:
gstack-timeline-stop (closes dangling session-timeline entries when a session
is interrupted; fail-open — 2s internal budget, always exits 0, can never block
a session). Skip it with ./setup --no-team, remove it with
gstack-settings-hook remove-source --source gstack-timeline-stop;
gstack-uninstall removes it too.
Hook registration is canonical-only: every hook command points at the stable
~/.claude/skills/gstack install, never the tree setup ran from, so deleting
a worktree or Conductor workspace can't leave dead hooks erroring in your
sessions. Every ./setup run also heals first: gstack-settings-hook prune-stale --repoint removes dead gstack hook entries, re-points stale ones
at the stable install, and collapses duplicates, printing one line (and
writing a backup beside the file) only when it changed something.
Continuous checkpoint mode (opt-in, local by default)
Set gstack-config set checkpoint_mode continuous and skills auto-commit your work as you go with a WIP: prefix plus a structured [gstack-context] body (decisions, remaining work, failed approaches). Survives crashes and context switches. /context-restore reads those commits to reconstruct session state. /ship filter-squashes WIP commits before the PR (preserving non-WIP commits) so bisect stays clean. Push is opt-in via checkpoint_push=true — default is local-only so you don't trigger CI on every WIP commit.
Domain skills + raw CDP escape hatch
Two new browser primitives compound the gstack agent over time:
$B domain-skill save— agent saves a per-site note (e.g., "LinkedIn's Apply button lives in an iframe") that fires automatically next time it visits that hostname. Quarantined → active after 3 successful uses → optional cross-project promotion via$B domain-skill promote-to-global. Storage lives alongside/learn's per-project learnings file. Full reference: docs/domain-skills.md.$B cdp <Domain.method>— raw Chrome DevTools Protocol escape hatch for the rare case curated commands miss. Deny-default: methods must be explicitly added tobrowse/src/cdp-allowlist.tswith a one-line justification. Two-tier mutex serializes browser-scoped CDP calls against per-tab work. Output for data-exfil methods is wrapped in the UNTRUSTED envelope.
Want raw CDP with no rails, no allowlist, no daemon — just thin transport from agent to Chrome? browser-use/browser-harness-js is a different philosophy (agent-authored helpers vs gstack's curated commands) and a good fit if you don't want gstack's security stack. The two can coexist: gstack's
$B cdpand harness can both attach to the same Chrome via Playwright'snewCDPSession.
Deep dives with examples and philosophy for every skill →
Karpathy's four failure modes? Already covered.
Andrej Karpathy's AI coding rules (17K stars) nail four failure modes: wrong assumptions, overcomplexity, orthogonal edits, imperative over declarative. gstack's workflow skills enforce all four. /office-hours forces assumptions into the open before code is written. The Confusion Protocol stops Claude from guessing on architectural decisions. /review catches unnecessary complexity and drive-by edits. /ship transforms tasks into verifiable goals with test-first execution. If you already use Karpathy-style CLAUDE.md rules, gstack is the workflow enforcement layer that makes them stick across entire sprints, not just single prompts.
Parallel sprints
gstack works well with one sprint. It gets interesting with ten running at once.
Design is at the heart. /design-consultation builds your design system from scratch, researches what's out there, proposes creative risks, and writes DESIGN.md. But the real magic is the shotgun-to-HTML pipeline.
/design-shotgun is how you explore. You describe what you want. It generates 4-6 AI mockup variants using GPT Image. Then it opens a comparison board in your browser with all variants side by side. You pick favorites, leave feedback ("more whitespace", "bolder headline", "lose the gradient"), and it generates a new round. Repeat until you love something. Taste memory kicks in after a few rounds so it starts biasing toward what you actually like. No more describing your vision in words and hoping the AI gets it. You see options, pick the good ones, and iterate visually.
/design-html makes it real. Take that approved mockup (from /design-shotgun, a CEO plan, a design review, or just a description) and turn it into production-quality HTML/CSS. Not the kind of AI HTML that looks fine at one viewport width and breaks everywhere else. This uses Pretext for computed text layout: text actually reflows on resize, heights adjust to content, layouts are dynamic. 30KB overhead, zero dependencies. It detects your framework (React, Svelte, Vue) and outputs the right format. Smart API routing picks different Pretext patterns depending on whether it's a landing page, dashboard, form, or card layout. The output is something you'd actually ship, not a demo.
/qa was a massive unlock. It let me go from 6 to 12 parallel workers. Claude Code saying "I SEE THE ISSUE" and then actually fixing it, generating a regression test, and verifying the fix — that changed how I work. The agent has eyes now.
Smart review routing. Just like at a well-run startup: CEO doesn't have to look at infra bug fixes, design review isn't needed for backend changes. gstack tracks what reviews are run, figures out what's appropriate, and just does the smart thing. The Review Readiness Dashboard tells you where you stand before you ship.
Test everything. /ship bootstraps test frameworks from scratch if your project doesn't have one. Every /ship run produces a coverage audit. Every /qa bug fix generates a regression test. 100% test coverage is the goal — tests make vibe coding safe instead of yolo coding.
/document-release is the engineer you never had. It reads every doc file in your project, cross-references the diff, and updates everything that drifted. README, ARCHITECTURE, CONTRIBUTING, CLAUDE.md, TODOS — all kept current automatically. And now /ship auto-invokes it — docs stay current without an extra command.
Real browser mode. /open-gstack-browser launches GStack Browser, an AI-controlled Chromium with anti-bot stealth, custom branding, and the sidebar extension baked in. Sites like Google and NYTimes work without captchas. The menu bar says "GStack Browser" instead of "Chrome for Testing." Your regular Chrome stays untouched. All existing browse commands work unchanged. $B disconnect returns to headless. The browser stays alive as long as the window is open... no idle timeout killing it while you're working.
Sidebar agent — your AI browser assistant. Type natural language in the Chrome side panel and a child Claude instance executes it. "Navigate to the settings page and screenshot it." "Fill out this form with test data." "Go through every item in this list and extract the prices." The sidebar auto-routes to the right model: Sonnet for fast actions (click, navigate, screenshot) and Opus for reading and analysis. Each task gets up to 5 minutes. The sidebar agent runs in an isolated session, so it won't interfere with your main Claude Code window. One-click cookie import right from the sidebar footer.
Personal automation. The sidebar agent isn't just for dev workflows. Example: "Browse my kid's school parent portal and add all the other parents' names, phone numbers, and photos to my Google Contacts." Two ways to get authenticated: (1) log in once in the headed browser, your session persists, or (2) click the "cookies" button in the sidebar footer to import cookies from your real Chrome. Once authenticated, Claude navigates the directory, extracts the data, and creates the contacts.
Prompt injection defense. Hostile web pages try to hijack your sidebar agent. gstack ships a layered defense: content filters (datamarking, hidden-element stripping, ARIA scrubbing, URL blocklist) on every page read, plus a 22MB ML classifier running locally in a sidecar subprocess that scans page-derived content before the agent sees it, with a verdict combiner that requires classifier agreement before blocking (prevents single-model false positives on Stack Overflow-style instruction pages). Everything runs on your machine, no network calls. Emergency kill switch: GSTACK_SECURITY_OFF=1. See ARCHITECTURE.md for the full stack.
Browser handoff when the AI gets stuck. Hit a CAPTCHA, auth wall, or MFA prompt? $B handoff opens a visible Chrome at the exact same page with all your cookies and tabs intact. Solve the problem, tell Claude you're done, $B resume picks up right where it left off. The agent even suggests it automatically after 3 consecutive failures.
/pair-agent is cross-agent coordination. You're in Claude Code. You also have OpenClaw running. Or Hermes. Or Codex. You want them both looking at the same website. Type /pair-agent, pick your agent, and a GStack Browser window opens so you can watch. The skill prints a block of instructions. Paste that block into the other agent's chat. It exchanges a one-time setup key for a session token, creates its own tab, and starts browsing. You see both agents working in the same browser, each in their own tab, neither able to interfere with the other. If ngrok is installed, the tunnel starts automatically so the other agent can be on a completely different machine. Same-machine agents get a zero-friction shortcut that writes credentials directly. This is the first time AI agents from different vendors can coordinate through a shared browser with real security: scoped tokens, tab isolation, rate limiting, domain restrictions, and activity attribution.
Multi-AI second opinion. /codex gets an independent review from OpenAI's Codex CLI — a completely different AI looking at the same diff. Three modes: code review with a pass/fail gate, adversarial challenge that actively tries to break your code, and open consultation with session continuity. When both /review (Claude) and /codex (OpenAI) have reviewed the same branch, you get a cross-model analysis showing which findings overlap and which are unique to each.
Safety guardrails on demand. Say "be careful" and /careful warns before any destructive command — rm -rf, DROP TABLE, force-push, git reset --hard. /freeze locks edits to one directory while debugging so Claude can't accidentally "fix" unrelated code. /guard activates both. /investigate auto-freezes to the module being investigated.
Proactive skill suggestions. gstack notices what stage you're in — brainstorming, reviewing, debugging, testing — and suggests the right skill. Don't like it? Say "stop suggesting" and it remembers across sessions.
10-15 parallel sprints
gstack is powerful with one sprint. It is transformative with ten running at once.
Conductor runs multiple Claude Code sessions in parallel — each in its own isolated workspace. One session running /office-hours on a new idea, another doing /review on a PR, a third implementing a feature, a fourth running /qa on staging, and six more on other branches. All at the same time. I regularly run 10-15 parallel sprints — that's the practical max right now.
The sprint structure is what makes parallelism work. Without a process, ten agents is ten sources of chaos. With a process — think, plan, build, review, test, ship — each agent knows exactly what to do and when to stop. You manage them the way a CEO manages a team: check in on the decisions that matter, let the rest run.
Voice input (AquaVoice, Whisper, etc.)
gstack skills have voice-friendly trigger phrases. Say what you want naturally — "run a security check", "test the website", "do an engineering review" — and the right skill activates. You don't need to remember slash command names or acronyms.
Uninstall
Option 1: Run the uninstall script
If gstack is installed on your machine:
~/.claude/skills/gstack/bin/gstack-uninstall
This handles skills, symlinks, global state (~/.gstack/), project-local state, browse daemons, and temp files. Use --keep-state to preserve config and analytics. Use --force to skip confirmation.
Option 2: Manual removal (no local repo)
If you don't have the repo cloned (e.g. you installed via a Claude Code paste and later deleted the clone):
# 1. Stop browse daemons
pkill -f "gstack.*browse" 2>/dev/null || true
# 2. Remove per-skill directories whose SKILL.md points into gstack/
# (rm -rf, not rmdir — installed dirs also contain runtime-asset links)
find ~/.claude/skills -mindepth 1 -maxdepth 1 -type d ! -name gstack 2>/dev/null |
while IFS= read -r dir; do
link="$dir/SKILL.md"
[ -L "$link" ] || continue
target=$(readlink "$link" 2>/dev/null) || continue
case "$target" in
gstack/*|*/gstack/*)
rm -rf "$dir"
;;
esac
done
# Alias skills install as copies (no symlink to detect) — remove by name
rm -rf ~/.claude/skills/_gstack-command ~/.claude/skills/connect-chrome 2>/dev/null
# 3. Remove gstack
rm -rf ~/.claude/skills/gstack
# 4. Remove global state
rm -rf ~/.gstack
# 5. Remove integrations (skip any you never installed)
rm -rf "${CODEX_HOME:-$HOME/.codex}/skills/gstack"* 2>/dev/null
rm -rf ~/.factory/skills/gstack* 2>/dev/null
rm -rf ~/.kiro/skills/gstack* 2>/dev/null
rm -rf ~/.openclaw/skills/gstack* 2>/dev/null
rm -rf ~/.cursor/skills/gstack* 2>/dev/null
rm -rf ~/.config/opencode/skills/gstack* 2>/dev/null
# 6. Remove temp files
rm -f /tmp/gstack-* 2>/dev/null
# 7. Per-project cleanup (run from each project root)
rm -rf .gstack .gstack-worktrees .claude/skills/gstack 2>/dev/null
rm -rf .agents/skills/gstack* .factory/skills/gstack* 2>/dev/null
Manual removal leaves gstack's hook entries behind in ~/.claude/settings.json
(the uninstall script removes all of them for you, including entries whose
_gstack_source tag was stripped). Edit that file and delete every hook whose
command path points into .claude/skills/gstack/: the SessionStart auto-update
hook, the AskUserQuestion PreToolUse/PostToolUse hooks, and the Stop hooks
(session timeline, plus verify-gate if you opted in). Left in place, they error
on every matching event once the install directory is gone.
Clean up CLAUDE.md
The uninstall script does not edit CLAUDE.md. In each project where gstack was added, remove the ## gstack and ## Skill routing sections.
Playwright
~/Library/Caches/ms-playwright/ (macOS) is left in place because other tools may share it. Remove it if nothing else needs it.
Free, MIT licensed, open source. No premium tier, no waitlist.
I open sourced how I build software. You can fork it and make it your own.
We're hiring. Want to ship real products at AI-coding speed and help harden gstack? Come work at YC — ycombinator.com/software Extremely competitive salary and equity. San Francisco, Dogpatch District.
GBrain — persistent knowledge for your coding agent
GBrain is a persistent knowledge base for AI agents — think of it as the memory your agent actually keeps between sessions. GStack gives you a one-command path from zero to "it's running, my agent can call it."
/setup-gbrain
Four paths, pick one:
- Supabase, existing URL — your cloud agent already provisioned a brain; paste the Session Pooler URL, now this laptop uses the same data.
- Supabase, auto-provision — paste a Supabase Personal Access Token; the skill creates a new project, polls to healthy, fetches the pooler URL, hands it to
gbrain init. ~90 seconds end-to-end. - PGLite local — zero accounts, zero network, ~30 seconds. Isolated brain on this Mac only. Great for try-first; migrate to Supabase later with
/setup-gbrain --switch. - Remote gbrain MCP — your brain runs on another machine (Tailscale, ngrok, internal LAN) or a teammate's server; paste an MCP URL and bearer token. Optionally pair with a local PGLite for symbol-aware code search in split-engine mode. Best for cross-machine memory without standing up a local DB.
After init, the skill offers to register gbrain as an MCP server for Claude Code (claude mcp add gbrain -- gbrain serve) so gbrain search, gbrain put, etc. show up as first-class typed tools — not bash shell-outs.
Keeping the brain current. Run /sync-gbrain from any repo to re-index its code into gbrain (incremental by default, --full for a full reindex, --dry-run to preview). The skill registers the cwd as a federated source via gbrain sources add, runs gbrain sync --strategy code, and writes a ## GBrain Search Guidance block to your project's CLAUDE.md so the agent prefers gbrain search/code-def/code-refs over Grep. The block is removed automatically if the capability check fails — no stale guidance pointing at tools that aren't installed.
Per-remote trust policy. Each repo on your machine gets one of three tiers:
read-write— agent can search the brain AND write new pages back from this reporead-only— agent can search but never writes (best for multi-client consultants: search the shared brain, don't contaminate it with Client A's work while in Client B's repo)deny— no gbrain interaction at all
The skill asks once per repo. The decision is sticky across worktrees and branches of the same remote.
GStack memory sync (different feature, same private-repo infra). Optionally pushes your gstack state (learnings, CEO plans, design docs, retros, developer profile) to a private git repo so your memory follows you across machines, with a one-time privacy prompt (everything allowlisted / artifacts only / off) and a defense-in-depth secret scanner that blocks AWS keys, tokens, PEM blocks, and JWTs before they leave your machine.
gstack-artifacts-init
Running gstack in Conductor? Conductor explicitly strips ANTHROPIC_API_KEY and OPENAI_API_KEY from every workspace's process env, so paid evals and gbrain embeddings won't work out of the box. Set GSTACK_ANTHROPIC_API_KEY and GSTACK_OPENAI_API_KEY in Conductor's workspace env config instead — gstack's TS entry points promote them to canonical names at runtime. Full details and the contributor checklist for adding the import to new entry points: Conductor + GSTACK_* env vars.
Full monty — every scenario, every flag, every bin helper, every troubleshooting step: USING_GBRAIN_WITH_GSTACK.md
Other references: docs/gbrain-sync.md (sync-specific guide) • docs/gbrain-sync-errors.md (error index)
Docs
| Doc | What it covers |
|---|---|
| Skill Deep Dives | Philosophy, examples, and workflow for every skill (includes Greptile integration) |
| Diagrams & Document Formats | Mermaid/excalidraw fences in PDFs, image sizing and safety defaults, --to html|docx, /diagram triplets |
| Builder Ethos | Builder philosophy: Boil the Ocean, Search Before Building, three layers of knowledge |
| Using GBrain with GStack | Every path, flag, bin helper, and troubleshooting step for /setup-gbrain |
| GBrain Sync | Cross-machine memory setup, privacy modes, troubleshooting |
| Architecture | Design decisions and system internals |
| Browser Reference | Full command reference for /browse |
| Contributing | Dev setup, testing, contributor mode, and dev mode |
| Changelog | What's new in every version |
Privacy & Telemetry
gstack includes opt-in usage telemetry to help improve the project. Here's exactly what happens:
- Default is off. Nothing is sent anywhere unless you explicitly say yes.
- On first run, gstack asks if you want to share anonymous usage data. You can say no.
- What's sent (if you opt in): skill name, duration, success/fail, gstack version, OS. That's it.
- What's never sent: code, file paths, repo names, branch names, prompts, or any user-generated content.
- Change anytime:
gstack-config set telemetry offdisables everything instantly. - Every off-machine send is receipted. Any gstack-initiated network send — telemetry included — writes a hash-chained, tamper-evident receipt to
~/.gstack/security/egress.jsonlbefore the send; sensitive sinks refuse to send at all if the receipt can't be written. Audit withgstack-egress list, verify the chain withgstack-egress verify(exit 3 on tamper), see the standing consent settings withgstack-egress grants. The ledger records attempted sends so accidents are auditable — it's an audit trail, not a network firewall.
Data is stored in Supabase (open source Firebase alternative). The schema is in supabase/migrations/ — you can verify exactly what's collected. The Supabase publishable key in the repo is a public key (like a Firebase API key) — row-level security policies deny all direct access. Telemetry flows through validated edge functions that enforce schema checks, event type allowlists, and field length limits.
Local analytics are always available. Run gstack-analytics to see your personal usage dashboard from the local JSONL file — no remote data needed.
Troubleshooting
Skill not showing up? cd ~/.claude/skills/gstack && ./setup
/browse fails? cd ~/.claude/skills/gstack && bun install && bun run build
Stale install? Run /gstack-upgrade — or set auto_upgrade: true in ~/.gstack/config.yaml
Want shorter commands? cd ~/.claude/skills/gstack && ./setup --no-prefix — switches from /gstack-qa to /qa. Your choice is remembered for future upgrades.
Want namespaced commands? cd ~/.claude/skills/gstack && ./setup --prefix — switches from /qa to /gstack-qa. Useful if you run other skill packs alongside gstack.
Codex says "Skipped loading skill(s) due to invalid SKILL.md"? Your Codex skill descriptions are stale. Fix: cd "${CODEX_HOME:-$HOME/.codex}/skills/gstack" && git pull && ./setup --host codex — or for repo-local installs: cd "$(readlink -f .agents/skills/gstack)" && git pull && ./setup --host codex
Windows users: gstack works on Windows 11 via Git Bash or WSL. Node.js is required in addition to Bun — Bun has a known bug with Playwright's pipe transport on Windows (bun#4253). The browse server automatically falls back to Node.js. Make sure both bun and node are on your PATH.
On Windows without Developer Mode (MSYS2 / Git Bash), setup falls back to file copies instead of symlinks because ln -snf produces frozen copies that don't refresh on git pull. Re-run cd ~/.claude/skills/gstack && ./setup after every git pull so your skill files match the repo. setup prints a one-line note reminding you. Unix and WSL keep symlinks and don't need the re-run.
Claude says it can't see the skills? Make sure your project's CLAUDE.md has a gstack section. Add this:
## gstack
Use /browse from gstack for all web browsing. Never use mcp__claude-in-chrome__* tools.
Available skills: /office-hours, /plan-ceo-review, /plan-eng-review, /plan-design-review,
/design-consultation, /design-shotgun, /design-html, /review, /ship, /land-and-deploy,
/canary, /benchmark, /browse, /open-gstack-browser, /qa, /qa-only, /design-review,
/setup-browser-cookies, /setup-deploy, /setup-gbrain, /sync-gbrain, /retro, /investigate,
/document-release, /document-generate, /codex, /cso, /autoplan, /pair-agent, /careful, /freeze,
/guard, /unfreeze, /gstack-upgrade, /learn.
License
MIT. Free forever. Go build something.