mirror of
https://github.com/garrytan/gstack.git
synced 2026-05-27 13:34:25 +02:00
920a13a17f463c28a3db75cc27482affb13a4fee
3 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Garry Tan
|
66f3a180d3 |
v1.43.2.0 fix wave: post-Daegu paper-cut — 18 fixes, 28 bisect commits (#1642)
* fix(gbrain-sync): --full produces an empty code index on first run of a new repo
`gbrain reindex-code` only RE-EMBEDS pages that already exist; it never walks
the filesystem. On a freshly-registered source (0 pages), a --full run that
called reindex-code alone found nothing ("No code pages to reindex"), finished
in ~1s, and left the code index permanently empty while still reporting OK.
Fix: --full now runs `sync --strategy code` FIRST to create pages via the file
walk, then runs `reindex-code` to honor the documented "full walk + reindex"
contract for both fresh and populated sources.
Contributed by @jetsetterfl via #1584.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(gbrain-local-status): classifier falsely reports broken-db inside repos with their own DATABASE_URL
The freshClassify probe ran `gbrain sources list --json` with the inherited
process env. When the probe ran from inside a repo with its own .env (an app
DATABASE_URL on a different port), Bun autoloaded the project's .env, gbrain
connected to the wrong database, and the classifier reported broken-db on
otherwise-healthy brains.
Fix: route the probe env through `buildGbrainEnv` from lib/gbrain-exec, the
same helper the sync orchestrator uses. DATABASE_URL is seeded from
~/.gbrain/config.json so the result is cwd-independent. The 60s cache can no
longer propagate a poisoned negative to clean directories.
Contributed by @jetsetterfl via #1583.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(retro): stale-base + bad-today-anchor pre-flight guard (#1624)
/retro silently produced confidently-wrong output when "today" drifted (model
session-context error) or when origin/<default> was materially behind the
actual remote — git log --since returned zero or near-zero commits and the
narrative was fabricated from nothing.
Adds Step 0.5 with four ordered pre-check branches before any window analysis:
A. No 'origin' remote → skip with "base freshness not verified" note
B. Detached HEAD → skip with "base freshness not verified" note
C. `git fetch origin <default>` fails (offline) → warn, proceed against
last-known origin/<default>
D. Fetch succeeded → compare today vs latest origin/<default> commit; if
gap > window-days, BLOCK with explicit citation of latest-commit date.
Skip paths still proceed to Step 1, but the disclosure is carried into the
retro narrative ("offline run, window not freshness-verified") so the output
is never silently confidently-wrong.
Atomic .tmpl + gen:skill-docs regen commit (T-Codex-3 pattern).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(retro): regression for #1624 stale-base pre-flight guard
13 static-invariant tests pinning the four ordered pre-check branches in
retro/SKILL.md.tmpl:Step 0.5:
A. no-remote skip — must check origin presence + set verdict
B. detached-HEAD skip — must gate behind prior verdict (ordering)
C. fetch-fail warn — must match `if !` or `||` shape, gate by verdict
D. stale-base BLOCK — must read latest-commit ISO date, cite remediation
Plus a disclosure-survives-to-narrative invariant: skip-path verdicts must be
named in prose so the retro output carries the cited reason rather than
silently misreporting.
Failing build if Step 0.5 is removed, branches re-ordered (no-remote no longer
wins), or the BLOCK message stops citing today/latest-commit/remediation
path.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(gbrain-sync): configurable timeouts + resume from gbrain checkpoint (#1611)
The memory and code stages hardcoded a 35-min spawn timeout. On brains with
~2000+ staged files, /sync-gbrain --full reliably SIGTERM'd the child at
exactly 35 minutes with exit 143. gbrain left ~/.gbrain/import-checkpoint.json
pointing at the staging dir, but gstack-memory-ingest's SIGTERM handler
unconditionally cleaned the dir up — so the next run found a checkpoint
pointing at nothing and restaged from scratch, repeating the SIGTERM forever.
Three changes:
1. Configurable timeouts via env (bounds 60_000ms - 86_400_000ms, default
2_100_000ms = 35min unchanged):
GSTACK_SYNC_MEMORY_TIMEOUT_MS
GSTACK_SYNC_CODE_TIMEOUT_MS
Out-of-range or non-numeric values warn and fall back to the default.
2. SIGTERM in gstack-memory-ingest no longer always cleans up the staging
dir. If gbrain has written ~/.gbrain/import-checkpoint.json pointing at
the active staging dir, the dir is PRESERVED for next-run resume.
Otherwise (no checkpoint pointing here, crash before gbrain ever
touched it) it's cleaned up as before.
3. Next /sync-gbrain run detects gbrain's checkpoint via decideResume() in
gstack-gbrain-sync.ts:
- no checkpoint → fresh ingest pass
- checkpoint + staging ok → set GSTACK_INGEST_RESUME_DIR; child
reuses staging dir and skips
writeStaged; gbrain import resumes
from processedIndex+1
- checkpoint + staging gone → warn "previous checkpoint stale
(staging dir gone), restaging from
scratch" and proceed
Reuses gbrain's own checkpoint as the source of truth (D1 — no double-store
state). Detect-then-fallback semantics per C1.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(gbrain-sync): regression for #1611 timeouts + resume
19 tests across three surfaces:
- resolveStageTimeoutMs (10 tests): undefined/empty → default; non-numeric,
zero, negative, below-floor, above-ceiling → warn + default; at-floor,
at-ceiling, valid mid-range → accepted as-is.
- decideResume (6 tests): no checkpoint, corrupt JSON, checkpoint + staging
ok, checkpoint + staging missing, checkpoint with no dir, checkpoint with
empty dir.
- SIGTERM staging preservation (3 static invariants): memory-ingest signal
handler must check stagingDirIsCheckpointed BEFORE cleanup; preserve
branch must come before cleanup branch (ordering); orchestrator must
pass GSTACK_INGEST_RESUME_DIR to the grandchild on resume.
Also threads process.env.HOME through readGbrainCheckpoint and
stagingDirIsCheckpointed so tests can redirect home. os.homedir() caches
at process start and ignores later mutation, so the env override is the
only reliable test injection point.
Failing build if the timeout bounds are removed, the resume detection
short-circuits incorrectly, or the SIGTERM handler regresses to
unconditional cleanup.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(review): pre-emit verification gate kills Django-shape FP class (#1539)
External user filed 4/8 false positives on a /review run against a Django +
DRF + PostgreSQL repo (Sprint 2.5). Every FP class was the same shape:
"resolvable in <5 minutes by viewing the actual code or running a simple
grep" — fields that don't exist on the model, dict.get()-might-be-None on a
form that returns {}-initialized cleaned_data, standard ORM save behavior
called out as data loss.
Extends the Confidence Calibration resolver (consumed by review, cso,
plan-eng-review, ship) with a Pre-emit verification gate:
Every finding MUST quote the specific code line that motivates it
(file:line + verbatim text). If the reviewer cannot produce the quote,
the finding is unverified — its confidence is forced to 4-5 so the
existing "Suppress from main report" rule fires automatically. The
finding still goes to the appendix for calibration audit, but the user
does not see it in the critical-pass output.
Reuses the existing suppression mechanism — no new code path. The FP
classes the gate kills are enumerated in the resolver text so reviewers
see the named patterns.
Framework-meta nudge included for Django Meta, Rails associations,
SQLAlchemy relationships, TypeORM decorators, Sequelize init, Prisma
generated client — the reviewer must quote the meta-construct that
generates the symbol, not just grep for the literal name. Deeper
framework-aware ORM verification (model introspection, migration-history-
aware checks) is deliberately deferred to a future wave per T-Codex-2.
Atomic .tmpl-equivalent (resolver) edit + gen:skill-docs regen commit
per T-Codex-3.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(review): regression for #1539 pre-emit verification gate
12 tests pinning the gate behavior:
- Resolver emits the gate header + #1539 reference
- Gate requires quoting file:line + verbatim text
- Unverified findings forced to confidence 4-5 (auto-suppress via
existing <7-rule, no new mechanism)
- Framework-meta nudge names Django, Rails, SQLAlchemy, TypeORM,
Sequelize, Prisma
- Deferred design doc reference present (1539-framework-aware-review.md)
- Four named FP classes from #1539 enumerated:
* field doesn't exist on model
* dict.get() might be None
* save() might lose fields
* update_fields might miss X
- All four downstream SKILL.md consumers (review, cso, plan-eng-review,
ship) carry the gate text after gen:skill-docs
- Existing confidence 9-10 'Show normally' + 3-4 'Suppress' rows
unchanged (regression on existing behavior)
Failing build if the gate is removed, the suppression mechanism is
re-invented separately, the framework-meta nudge drops a framework, or
gen:skill-docs stops propagating the gate to consumers.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(config): expose explain_level default
* fix(benchmark): parse positional prompt after flags
* fix(artifacts): reject malformed remote paths
* fix(learnings): preserve current entries in cross-project search
* fix(setup): register root gstack slash alias
* fix(memory): probe gitleaks without shell builtin
* fix(gbrain-lib): pin LC_ALL=C in varname validator (macOS locale guard)
In many macOS shells the default locale (e.g. en_US.UTF-8) makes bash
glob brackets like `[A-Z]` match lowercase letters too, so the existing
`case "$name" in [A-Z_][A-Z0-9_]*)` branch lets names like `lower-case`
through validation. The function then trips `printf -v "$varname"` and
`export "$varname"` with `not a valid identifier` errors that surface
mid-prompt, which is exactly what the validator was supposed to prevent.
Pinning `LC_ALL=C` inside the function gives ASCII-only bracket semantics
on both macOS and Linux, matching the documented `[A-Z_][A-Z0-9_]*`
contract. Declared `local` so it doesn't leak to the calling shell —
`gstack-gbrain-lib.sh` is documented as a sourced helper, so a bare
assignment would mutate the caller's locale for the rest of the process
(silently affecting downstream `sort`, `tr`, locale-aware globs in the
same shell, etc.).
The existing regression test
`test/gbrain-lib-verify.test.ts:'rejects invalid var names'`
already covers the macOS repro shape (passes `lower-case` and expects
the validator to reject + emit `invalid var name`). On Linux CI the
test silently passed because `LC_ALL=C` is the typical default; on
macOS dev boxes it fails.
Verified:
- `bun test test/gbrain-lib-verify.test.ts`: 22 pass, 0 fail (on macOS).
- `_gstack_gbrain_validate_varname lower-case; echo $?` → 2.
- `_gstack_gbrain_validate_varname FOO_BAR; echo $?` → 0.
- Caller's LC_ALL preserved across calls (confirmed via sourced bash).
* fix(land-and-deploy): detect merged PR after gh failure
After `gh pr merge` exits non-zero, the PR may already be MERGED server-side
(concurrent merge landed, or local cleanup phase failed AFTER the merge
succeeded). Calling `gh pr merge` a second time then errors with a confusing
"already merged" — and worse, the deploy workflow never runs because we
stopped on the first failure.
Adds a Post-failure PR-state check (§4a-postfail) that runs after ANY
non-zero exit from `gh pr merge`:
- state == MERGED → record MERGE_PATH=direct, OFFER (don't force)
stale-worktree cleanup on the base branch with
uncommitted-work guard, proceed to §4a CI watch
- state == OPEN → check autoMergeRequest; if non-null treat as
merge-queue wait; if null surface both errors and STOP
- state == CLOSED → STOP
Hard invariant: never retry `gh pr merge` after a non-zero exit. Server
state is authoritative.
Re-authored from PR #1620 into land-and-deploy/SKILL.md.tmpl (the source of
truth) instead of the generated SKILL.md, so the next gen:skill-docs run
preserves the change. Original diff by @davidfoy via #1620.
Related: cli/cli#3442, cli/cli#13380.
Contributed by @davidfoy via #1620.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix: detect PgBouncer transaction-mode pooler and set GBRAIN_PREPARE=true (#1435)
When gbrain connects through a PgBouncer transaction-mode pooler (port
6543), it auto-disables prepared statements. This breaks `gbrain search`
silently — the /sync-gbrain capability check fails and the GBrain Search
Guidance block never gets written to CLAUDE.md.
Three-layer fix:
1. **lib/gbrain-exec.ts** — `buildGbrainEnv()` now detects port 6543 in
the effective DATABASE_URL and sets `GBRAIN_PREPARE=true` in the env
passed to every gbrain spawn. This is the single chokepoint — all
gstack gbrain invocations inherit the fix. Caller can opt out with
`GBRAIN_PREPARE=false`.
2. **sync-gbrain/SKILL.md{,.tmpl}** — capability check now exports
`GBRAIN_PREPARE=true` explicitly and retries search up to 3x with 1s
delay for async index propagation under connection pooling.
3. **bin/gstack-gbrain-detect** — surfaces `gbrain_pooler_mode` field
("transaction" | "session" | null) in the preamble probe JSON so
/setup-gbrain and /sync-gbrain can advise users about pooler state.
Closes #1435
Built with [ClosedLoop.AI](https://closedloop.ai) | [GitHub](https://github.com/closedloop-ai/claude-plugins)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(supabase-provision): rewrite transaction/6543 -> session/5432 for new projects
- Single-object pooler API responses default to transaction-mode at 6543,
but the shared pooler tenant on new projects only listens on session/5432
- Add a `pool_mode == transaction && db_port == 6543` rewrite + stderr note
- Escape hatch via `GSTACK_SUPABASE_TRUST_API_PORT=1` for forward-compat
- 5 new tests covering rewrite, no-op shapes, env opt-out, array path
Fixes #1301.
* fix(browse): GSTACK_CHROMIUM_NO_SANDBOX opt-out for Ubuntu/AppArmor (#1562)
Ubuntu/AppArmor configurations often block unprivileged Chromium sandboxing
for headless agent sessions even for normal users — /qa hangs without
--no-sandbox. The kernel policy denies the unprivileged user namespaces
Chromium needs.
Adds GSTACK_CHROMIUM_NO_SANDBOX=1 as an explicit user override that forces
the sandbox off without changing the default for everyone else. Re-authored
from PR #1562 onto v1.42.2.0's shouldEnableChromiumSandbox() helper —
purely additive, preserves the headed-launch sandbox-on-by-default behavior
that v1.42.2.0 shipped to kill the --no-sandbox yellow infobar.
Three new regression tests cover:
- linux + override=1 → false (the named use case)
- darwin + override=1 → false (env wins on any platform)
- override=0 → does NOT trigger (must be exactly "1")
Original diff by @techcenter68 via #1562.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(browse): mirror isCustomChromium() guard in headless launch()
When BROWSE_EXTENSIONS_DIR is set alongside GSTACK_CHROMIUM_PATH pointing
at a baked-extension build (GBrowser / GStack Browser), the headless launch()
path was unconditionally adding --disable-extensions-except / --load-extension.
This causes the same ServiceWorkerState::SetWorkerId DCHECK crash that
launchHeaded() already guards against via isCustomChromium().
Mirror the existing guard: skip --load-extension flags when isCustomChromium()
returns true; always push the off-screen window geometry args.
* fix(browse): daemonize macOS/Linux server via setsid()
`Bun.spawn().unref()` only releases the child from Bun's event loop —
it does NOT call setsid(). The spawned bun server inherits the spawning
shell's process session. When the CLI runs inside a session-managed shell
that exits shortly after the CLI returns (Claude Code's per-command Bash
sandbox, Conductor, OpenClaw, CI step runners), the session leader's exit
sends SIGHUP to every PID in the session — killing the bun server and
its Chromium grandchildren within seconds of a successful `connect`.
Setting `BROWSE_PARENT_PID=0` (already done by the `connect` command and
pair-agent) disables the parent-process watchdog but does NOT save the
server here: SIGHUP from session teardown still reaps it.
Replace the macOS/Linux `Bun.spawn().unref()` with Node's
`child_process.spawn({ detached: true })`, which calls setsid() and
gives the server its own session leader role (PPID=1, STAT=Ss). This
mirrors the Windows path's rationale (PR #191 by @fqueiro) — same root
cause, different OS surface.
Verified on macOS in Conductor: pre-fix the server dies ~10–15s after
connect across separate Bash invocations; post-fix the same PID stays
alive (PPID=1, SESS=0, STAT=Ss) and responds to `status`/`goto`/
`snapshot` across many separate shell calls.
The `proc?.stderr` startup-error branch is removed since both platforms
now spawn with `stdio: 'ignore'`; both fall through to the on-disk
`browse-startup-error.log` written by `server.ts`'s start().catch.
* fix(design): bump image-gen timeout to 240s + pin gpt-image-2
The design binary calls /v1/responses (gpt-4o + image_generation tool,
quality:high, 1536x1024) but aborted the request after a hardcoded 120s.
That class of request consistently takes ~140-160s end-to-end, so every
generate/variants/evolve/iterate call aborted before the image returned.
In /design-shotgun this cascades: Step 3c launches N parallel agents,
each calling `$D generate`, each aborts at 120s and retries, all fail,
the comparison board never opens — the skill appears to hang indefinitely.
Reproduced the exact API call with a longer budget: HTTP 200, valid
image, 143.5s. A real /design-shotgun run after the patch generated 3
variants in parallel at 150.0s / 161.0s / 152.1s, all exit 0 — note the
161s case, which a naive 150s bump would still have failed.
- Bump AbortController timeout 120_000 -> 240_000 in generate.ts,
variants.ts, evolve.ts, iterate.ts (both call sites)
- Pin the image_generation tool to model "gpt-image-2"
design/test/variants-retry-after.test.ts: 5 pass, 0 fail. The
feedback-roundtrip.test.ts failures are a pre-existing browse-module
breakage (session.clearLoadedHtml undefined), unrelated to this change.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test: fill coverage gaps for PRs #1606, #1612, #1620
Three cherry-picked PRs in this wave landed without unit-test coverage for
the specific invariant they protect:
#1606 (@andrey-esipov) — LC_ALL=C pin in _gstack_gbrain_validate_varname
8 tests by sourcing bin/gstack-gbrain-lib.sh and calling the validator
directly. Asserts uppercase/digit/underscore accepted, lowercase
REJECTED (the macOS-locale regression case), mixed-case rejected,
LC_ALL=C scoping is local (doesn't leak to caller).
#1612 (@bharat2913) — setsid daemonize via Node child_process.spawn
4 static-invariant tests on browse/src/cli.ts. The actual setsid
syscall is hard to assert without a real spawn, so we pin the source
shape: nodeSpawn imported from child_process; non-Windows branch uses
nodeSpawn(...) with detached:true and .unref(); comment documents
setsid/SIGHUP root cause; Bun.spawn() is NOT used on macOS/Linux.
#1620 (@davidfoy, re-authored into .tmpl per A3) — §4a-postfail
12 static invariants on land-and-deploy/SKILL.md.tmpl + generated
SKILL.md. Pins all three state branches (MERGED/OPEN/CLOSED), the
authoritative state query, the merge-SHA capture, non-destructive
worktree cleanup with uncommitted-work guard, autoMergeRequest probe
on OPEN, hard "never retry gh pr merge" rule, and atomic regen
propagation.
Failing build if any of the three invariants regresses.
Note: gbrain-lib-validate-varname.test.ts also surfaces a pre-existing
glob-pattern overpermissiveness (hyphens + dots accepted) — not in
#1606's scope; documented inline as a separate cleanup target.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(learnings): align injection-prevention tests with PR #1619 tagged-line shape
PR #1619 (preserve current entries in cross-project search) refactored
gstack-learnings-search to tag rows inline (`current\t<json>` vs
`cross\t<json>`) instead of filtering inside the bun block via
process.env.GSTACK_SEARCH_SLUG. The bun block no longer reads SLUG or
CROSS env vars — it parses the per-line tag and sets a per-entry
_crossProject flag.
The pre-existing test/learnings-injection.test.ts still asserted on the
old SLUG + CROSS env var shape. Updates:
- Remove the SLUG env var assertion (no longer set on bash command line)
- Remove the bun-block CROSS env var assertion (block reads the tag now,
not the env)
- Add a new positive assertion that the bun block parses the tag
(sourceTag | tabIndex | crossProject)
- Keep the shell-interpolation safety assertion unchanged — that's
independent of the SLUG refactor
The CROSS env var is still SET on the bash command line (it controls
whether the cross-project find runs at all), but the bun child no longer
reads it. The existing "env vars set on bash command line" test continues
to pin that.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(fixtures): regenerate ship-SKILL.md golden baselines
ship/SKILL.md consumes the Confidence Calibration resolver via the
preamble pipeline. This wave's #1539 pre-emit verification gate extends
the resolver text, which propagated to ship/SKILL.md via gen:skill-docs.
The golden fixtures in test/fixtures/golden/ matched the pre-#1539 shape
and failed the host-config regression check.
Refreshes claude-ship-SKILL.md, codex-ship-SKILL.md, and factory-ship-SKILL.md
to match the current generated output. Matches the Daegu wave's bisect
commit 23 ("test(fixtures): regenerate ship-SKILL.md golden baselines").
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test(gbrain-detect): include gbrain_pooler_mode in schema regression (PR #1591)
PR #1591 (PgBouncer transaction-mode detection, @mikeangstadt) added
gbrain_pooler_mode to the gstack-gbrain-detect JSON output but did not
update the schema regression check in
test/gstack-gbrain-detect-mcp-mode.test.ts. Adding the key in alphabetical
order matching the rest of the schema array. Downstream sync-gbrain ignores
unknown keys, so this is forward-compat.
Without this, the test fails with a diff:
+ "gbrain_pooler_mode"
because keys is the actual set returned and the expected array was
pre-#1591.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore(release): v1.43.0.0 — post-Daegu paper-cut wave
Bumps VERSION 1.42.2.0 → 1.43.0.0 (MINOR per scale-aware bump rules: new
env-var surface GSTACK_SYNC_*_TIMEOUT_MS + GSTACK_CHROMIUM_NO_SANDBOX,
behavior expansion in browse/src/browser-manager.ts headless launch,
three skill-template prompt changes affecting /retro, /review,
/sync-gbrain).
CHANGELOG entry leads with what stopped happening: /retro stops
fabricating retros against stale bases, /sync-gbrain stops SIGTERM-looping
35-min restarts on big brains, /review stops shipping framework FPs the
reviewer never grep'd.
18 fixes total — 15 community PRs + 3 self-filed silent-failure issues
(#1624, #1611, #1539) — in one bundled PR with 26 bisect commits and 7
new regression test files. Every wave-touched test file passes in
isolation.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore(release): bump v1.43.0.0 → v1.43.2.0 for queue collision
CI check-version-stale flagged v1.43.0.0 already claimed by PR #1574
(garrytan/colombo-v3). PR #1639 (garrytan/muscat-v3) claims v1.43.1.0.
Next available MINOR slot is v1.43.2.0.
Bump VERSION + package.json + CHANGELOG entry header. No behavior
changes — purely re-versioning to clear the queue collision.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-authored-by: Jayesh Betala <jayesh.betala7@gmail.com>
Co-authored-by: Andrey Esipov <andrey.esipov@outlook.com>
Co-authored-by: David Foy <davidfoy@users.noreply.github.com>
Co-authored-by: mikeangstadt <mike.angstadt@closedloop.ai>
Co-authored-by: 0xDevNinja <manmit0x@gmail.com>
Co-authored-by: techcenter68 <techcenter68@users.noreply.github.com>
Co-authored-by: shohu <shohu33@gmail.com>
Co-authored-by: Bharat <bharat@theysaid.io>
Co-authored-by: Matteo Hertel <info@matteohertel.com>
|
||
|
Garry Tan
|
33cb4715ef |
v1.39.2.0 feat: GSTACK_* env-shim for Conductor + gbrain/gstack setup docs (#1534)
* feat: GSTACK_* env-key shim for Conductor workspaces New lib/conductor-env-shim.ts promotes GSTACK_ANTHROPIC_API_KEY and GSTACK_OPENAI_API_KEY to canonical names when canonical is empty. Wired into the four TS entry points that hit paid APIs or gbrain embeddings: gstack-gbrain-sync.ts, gstack-model-benchmark, preflight-agent-sdk.ts, test/helpers/e2e-helpers.ts. Side-effect-only import, 15 lines total. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * docs: gbrain+gstack setup, Conductor env mapping (v1.39.2.0) USING_GBRAIN_WITH_GSTACK.md: new "What you get after setup" section, Path 4 (remote MCP / split-engine), /sync-gbrain workflow stages + watermark mechanics, "Conductor + GSTACK_* env vars" section, env vars table extended, two troubleshooting entries (silent embedding failure and FILE_TOO_LARGE watermark block). CONTRIBUTING.md "Conductor workspaces": new paragraph on the GSTACK_* prefix pattern and the four entry points importing the shim. VERSION 1.39.1.0 → 1.39.2.0 and CHANGELOG entry covering the shim + docs (full release-summary format with before/after table). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * test: unit coverage for conductor-env-shim Refactor lib/conductor-env-shim.ts to export promoteConductorEnv() so unit tests can manipulate env and call it directly (a bare side- effect IIFE on import isn't reachable from bun:test once cached). The on-import IIFE still runs — existing four-entry-point imports keep working unchanged. test/conductor-env-shim.test.ts covers all three branches: GSTACK_FOO present + FOO empty → promotion; FOO already set → no-overwrite; nothing in env → no-op. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * docs: Conductor strips canonical API keys (not just "doesn't inherit") The prior docs framed the GSTACK_* prefix as collision-avoidance: "Conductor exposes API keys under a GSTACK_ prefix so it never collides with whatever the host system has set." That understates the mechanism — Conductor actively strips ANTHROPIC_API_KEY and OPENAI_API_KEY from every workspace's process env, so setting them in ~/.zshrc or .env doesn't help. The fix path is to set the GSTACK_-prefixed forms in Conductor's workspace env config; Conductor passes those through untouched. Three docs updated to reflect the strip, not the polite framing: USING_GBRAIN_WITH_GSTACK.md (Conductor section), CONTRIBUTING.md (Conductor workspaces paragraph), CHANGELOG.md (release summary). README.md gains a "Running gstack in Conductor?" callout in the GBrain section pointing at the canonical doc's anchor, plus a fourth path entry (remote gbrain MCP / split-engine) that was already documented in USING_GBRAIN but missing from the README summary. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com> |
||
|
Garry Tan
|
22a4451e0e |
feat(v1.3.0.0): open agents learnings + cross-model benchmark skill (#1040)
* chore: regenerate stale ship golden fixtures
Golden fixtures were missing the VENDORED_GSTACK preamble section that
landed on main. Regression tests failed on all three hosts (claude, codex,
factory). Regenerated from current preamble output.
No code changes, unblocks test suite.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat: anti-slop design constraints + delete duplicate constants
Tightens design-consultation and design-shotgun to push back on the
convergence traps every AI design tool falls into.
Changes:
- scripts/resolvers/constants.ts: add "system-ui as primary font" to
AI_SLOP_BLACKLIST. Document Space Grotesk as the new "safe alternative
to Inter" convergence trap alongside the existing overused fonts.
- scripts/gen-skill-docs.ts: delete duplicate AI slop constants block
(dead code — scripts/resolvers/constants.ts is the live source).
Prevents drift between the two definitions.
- design-consultation/SKILL.md.tmpl: add Space Grotesk + system-ui to
overused/slop lists. Add "anti-convergence directive" — vary across
generations in the same project. Add Phase 1 "memorable-thing forcing
question" (what's the one thing someone will remember?). Add Phase 5
"would a human designer be embarrassed by this?" self-gate before
presenting variants.
- design-shotgun/SKILL.md.tmpl: anti-convergence directive — each
variant must use a different font, palette, and layout. If two
variants look like siblings, one of them failed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat: context health soft directive in preamble (T2+)
Adds a "periodically self-summarize" nudge to long-running skills.
Soft directive only — no thresholds, no enforcement, no auto-commit.
Goal: self-awareness during /qa, /investigate, /cso etc. If you notice
yourself going in circles, STOP and reassess instead of thrashing.
Codex review caught that fake precision thresholds (15/30/45 tool calls)
were unimplementable — SKILL.md is a static prompt, not runtime code.
This ships the soft version only.
Changes:
- scripts/resolvers/preamble.ts: add generateContextHealth(), wire into
T2+ tier. Format: [PROGRESS] ... summary line. Explicit rule that
progress reporting must never mutate git state.
- All T2+ skill SKILL.md files regenerated to include the new section.
- Golden ship fixtures updated (T4 skill, picks up the change).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat: model overlays with explicit --model flag (no auto-detect)
Adds a per-model behavioral patch layer orthogonal to the host axis.
Different LLMs have different tendencies (GPT won't stop, Gemini
over-explains, o-series wants structured output). Overlays nudge each
model toward better defaults for gstack workflows.
Codex review caught three landmines the prior reviews missed:
1. Host != model — Claude Code can run any Claude model, Codex runs
GPT/o-series, Cursor fronts multiple providers. Auto-detecting from
host would lie. Dropped auto-detect. --model is explicit (default
claude). Missing overlay file → empty string (graceful).
2. Import cycle — putting Model in resolvers/types.ts would cycle
through hosts/index. Created neutral scripts/models.ts instead.
3. "Final say" is dangerous — overlay at the end of preamble could
override STOP points, AskUserQuestion gates, /ship review gates.
Placed overlay after spawned-session-check but before voice + tier
sections. Wrapper heading adds explicit subordination language on
every overlay: "subordinate to skill workflow, STOP points,
AskUserQuestion gates, plan-mode safety, and /ship review gates."
Changes:
- scripts/models.ts: new neutral module. ALL_MODEL_NAMES, Model type,
resolveModel() for family heuristics (gpt-5.4-mini → gpt-5.4, o3 →
o-series, claude-opus-4-7 → claude), validateModel() helper.
- scripts/resolvers/types.ts: import Model, add ctx.model field.
- scripts/resolvers/model-overlay.ts: new resolver. Reads
model-overlays/{model}.md. Supports {{INHERIT:base}} directive at
top of file for concat (gpt-5.4 inherits gpt). Cycle guard.
- scripts/resolvers/index.ts: register MODEL_OVERLAY resolver.
- scripts/resolvers/preamble.ts: wire generateModelOverlay into
composition before voice. Print MODEL_OVERLAY: {model} in preamble
bash so users can see which overlay is active. Filter empty sections.
- scripts/gen-skill-docs.ts: parse --model CLI flag. Default claude.
Unknown model → throw with list of valid options.
- model-overlays/{claude,gpt,gpt-5.4,gemini,o-series}.md: behavioral
patches per model family. gpt-5.4.md uses {{INHERIT:gpt}} to extend
gpt.md without duplication.
- test/gen-skill-docs.test.ts: fix qa-only guardrail regex scope.
Was matching Edit/Glob/Grep anywhere after `allowed-tools:` in the
whole file. Now scoped to frontmatter only. Body prose (Claude
overlay references Edit as a tool) correctly no longer breaks it.
Verification:
- bun run gen:skill-docs --host all --dry-run → all fresh
- bun run gen:skill-docs --model gpt-5.4 → concat works, gpt.md +
gpt-5.4.md content appears in order
- bun run gen:skill-docs --model unknown → errors with valid list
- All generated skills contain MODEL_OVERLAY: claude in preamble
- Golden ship fixtures regenerated
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat: continuous checkpoint mode with non-destructive WIP squash
Adds opt-in auto-commit during long sessions so work survives Claude
Code crashes, Conductor workspace handoffs, and context switches.
Local-only by default — pushing requires explicit opt-in.
Codex review caught multiple landmines that would have shipped:
1. checkpoint_push=true default would push WIP commits to shared
branches, trigger CI/deploys, expose secrets. Now default false.
2. Plan's original /ship squash (git reset --soft to merge base) was
destructive — uncommitted ALL branch commits, not just WIP, and
caused non-fast-forward pushes. Redesigned: rebase --autosquash
scoped to WIP commits only, with explicit fallback for WIP-only
branches and STOP-and-ask for conflicts.
3. gstack-config get returned empty for missing keys with exit 0,
ignoring the annotated defaults in the header comments. Fixed:
get now falls back to a lookup_default() table that is the
canonical source for defaults.
4. Telemetry default mismatched: header said 'anonymous' but runtime
treated empty as 'off'. Aligned: default is 'off' everywhere.
5. /checkpoint resume only read markdown checkpoint files, not the
WIP commit [gstack-context] bodies the plan referenced. Wired up
parsing of [gstack-context] blocks from WIP commits as a second
recovery trail alongside the markdown checkpoints.
Changes:
- bin/gstack-config: add checkpoint_mode (default explicit) and
checkpoint_push (default false) to CONFIG_HEADER. Add lookup_default()
as canonical default source. get() falls back to defaults when key
absent. list now shows value + source (set/default). New 'defaults'
subcommand to inspect the table.
- scripts/resolvers/preamble.ts: preamble bash reads _CHECKPOINT_MODE
and _CHECKPOINT_PUSH, prints CHECKPOINT_MODE: and CHECKPOINT_PUSH: so
the mode is visible. New generateContinuousCheckpoint() section in
T2+ tier describes WIP commit format with [gstack-context] body and
the rules (never git add -A, never commit broken tests, push only
if opted in). Example deliberately shows a clean-state context so
it doesn't contradict the rules.
- ship/SKILL.md.tmpl: new Step 5.75 WIP Commit Squash. Detects WIP
count, exports [gstack-context] blocks before squash (as backup),
uses rebase --autosquash for mixed branches and soft-reset only when
VERIFIED WIP-only. Explicit anti-footgun rules against blind soft-
reset. Aborts with BLOCKED status on conflict instead of destroying
non-WIP commits.
- checkpoint/SKILL.md.tmpl: new Step 1.5 to parse [gstack-context]
blocks from WIP commits via git log --grep="^WIP:". Merges with
markdown checkpoint for fuller session recovery.
- Golden ship fixtures regenerated (ship is T4, preamble change shows up).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat: feature discovery flow gated by per-feature markers
Extends generateUpgradeCheck() to surface new features once per user
after a just-upgraded session. No more silent features.
Codex review caught: spawned sessions (OpenClaw, etc.) must skip the
discovery prompt entirely — they can't interactively answer. Feature
discovery now checks SPAWNED_SESSION first and is silent in those.
Discovery is per-feature, not per-upgrade. Each feature has its own
marker file at ~/.claude/skills/gstack/.feature-prompted-{name}. Once
the user has been shown a feature (accepted, shown docs, or skipped),
the marker is touched and the prompt never fires again for that
feature. Future features get their own markers.
V1 features surfaced:
- continuous-checkpoint: offer to enable checkpoint_mode=continuous
- model-overlay: inform-only note about --model flag and MODEL_OVERLAY
line in preamble output
Max one prompt per session to avoid nagging. Fires only on JUST_UPGRADED
(not every session), plus spawned-session skip.
Changes:
- scripts/resolvers/preamble.ts: extend generateUpgradeCheck() with
feature discovery rules, per-marker-file semantics, spawned-session
exclusion, and max-one-per-session cap.
- All skill SKILL.md files regenerated to include the new section.
- Golden ship fixtures regenerated.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat: design taste engine with persistent schema
Adds a cross-session taste profile that learns from design-shotgun
approval/rejection decisions. Biases future design-consultation and
design-shotgun proposals toward the user's demonstrated preferences.
Codex review caught that the plan had "taste engine" as a vague goal
without schema, decay, migration, or placeholder insertion points. This
commit ships the full spec.
Schema v1 at ~/.gstack/projects/$SLUG/taste-profile.json:
- version, updated_at
- dimensions: fonts, colors, layouts, aesthetics — each with approved[]
and rejected[] preference lists
- sessions: last 50 (FIFO truncation), each with ts/action/variant/reason
- Preference: { value, confidence, approved_count, rejected_count, last_seen }
- Confidence: Laplace-smoothed approved/(total+1)
- Decay: 5% per week of inactivity, computed at read time (not write)
Changes:
- bin/gstack-taste-update: new CLI. Subcommands approved/rejected/show/
migrate. Parses reason string for dimension signals (e.g.,
"fonts: Geist; colors: slate; aesthetics: minimal"). Emits taste-drift
NOTE when a new signal contradicts a strong opposing signal. Legacy
approved.json aggregates migrate to v1 on next write.
- scripts/resolvers/design.ts: new generateTasteProfile() resolver.
Produces the prose that skills see: how to read the profile, how to
factor into proposals, conflict handling, schema migration.
- scripts/resolvers/index.ts: register TASTE_PROFILE and a BIN_DIR
resolver (returns ctx.paths.binDir, used by templates that shell out
to gstack-* binaries).
- design-consultation/SKILL.md.tmpl: insert {{TASTE_PROFILE}} placeholder
in Phase 1 right after the memorable-thing forcing question so the
Phase 3 proposal can factor in learned preferences.
- design-shotgun/SKILL.md.tmpl: taste memory section now reads
taste-profile.json via {{TASTE_PROFILE}}, falls back to per-session
approved.json (legacy). Approval flow documented to call
gstack-taste-update after user picks/rejects a variant.
Known gap: v1 extracts dimension signals from a reason string passed
by the caller ("fonts: X; colors: Y"). Future v2 can read EXIF or an
accompanying manifest written by design-shotgun alongside each variant
for automatic dimension extraction without needing the reason argument.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat: multi-provider model benchmark (boil the ocean)
Adds the full spec Codex asked for: real provider adapters with auth
detection, normalized RunResult, pricing tables, tool compatibility
maps, parallel execution with error isolation, and table/JSON/markdown
output. Judge stays on Anthropic SDK as the single stable source of
quality scoring, gated behind --judge.
Codex flagged the original plan as massively under-scoped — the
existing runner is Claude-only and the judge is Anthropic-only. You
can't benchmark GPT or Gemini without real provider infrastructure.
This commit ships it.
New architecture:
test/helpers/providers/types.ts ProviderAdapter interface
test/helpers/providers/claude.ts wraps `claude -p --output-format json`
test/helpers/providers/gpt.ts wraps `codex exec --json`
test/helpers/providers/gemini.ts wraps `gemini -p --output-format stream-json --yolo`
test/helpers/pricing.ts per-model USD cost tables (quarterly)
test/helpers/tool-map.ts which tools each CLI exposes
test/helpers/benchmark-runner.ts orchestrator (Promise.allSettled)
test/helpers/benchmark-judge.ts Anthropic SDK quality scorer
bin/gstack-model-benchmark CLI entry
test/benchmark-runner.test.ts 9 unit tests (cost math, formatters, tool-map)
Per-provider error isolation:
- auth → record reason, don't abort batch
- timeout → record reason, don't abort batch
- rate_limit → record reason, don't abort batch
- binary_missing → record in available() check, skip if --skip-unavailable
Pricing correction: cached input tokens are disjoint from uncached
input tokens (Anthropic/OpenAI report them separately). Original
math subtracted them, producing negative costs. Now adds cached at
the 10% discount alongside the full uncached input cost.
CLI:
gstack-model-benchmark --prompt "..." --models claude,gpt,gemini
gstack-model-benchmark ./prompt.txt --output json --judge
gstack-model-benchmark ./prompt.txt --models claude --timeout-ms 60000
Output formats: table (default), json, markdown. Each shows model,
latency, in→out tokens, cost, quality (when --judge used), tool calls,
and any errors.
Known limitations for v1:
- Claude adapter approximates toolCalls as num_turns (stream-json
would give exact counts; v2 can upgrade).
- Live E2E tests (test/providers.e2e.test.ts) not included — they
require CI secrets for all three providers. Unit tests cover the
shape and math.
- Provider CLIs sometimes return non-JSON error text to stdout; the
parsers fall back to treating raw output as plain text in that case.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat: standalone methodology skill publishing via gstack-publish
Ships the marketplace-distribution half of Item 5 (reframed): publish
the existing standalone OpenClaw methodology skills to multiple
marketplaces with one command.
Codex review caught that the original plan assumed raw generated
multi-host skills could be published directly. They can't — those
depend on gstack binaries, generated host paths, tool names, and
telemetry. The correct artifact class is hand-crafted standalone
skills in openclaw/skills/gstack-openclaw-* (already exist and work
without gstack runtime). This commit adds the wrapper that publishes
them to ClawHub + SkillsMP + Vercel Skills.sh with per-marketplace
error isolation and dry-run validation.
Changes:
- skills.json: root manifest with 4 skills (office-hours, ceo-review,
investigate, retro) each pointing at its openclaw/skills source.
Each skill declares per-marketplace targets with a slug, a publish
flag, and a compatible-hosts list. Marketplace configs include CLI
name, login command, publish command template (with placeholder
substitution), docs URL, and auth_check command.
- bin/gstack-publish: new CLI. Subcommands:
gstack-publish Publish all skills
gstack-publish <slug> Publish one skill
gstack-publish --dry-run Validate + auth-check without publishing
gstack-publish --list List skills + marketplace targets
Features:
* Manifest validation (missing source files, missing slugs, empty
marketplace list all reported).
* Per-marketplace auth check before any publish attempt.
* Per-skill / per-marketplace error isolation: one failure doesn't
abort the batch.
* Idempotent — re-running with the same version is safe; markets
that reject duplicate versions report it as a failure for that
single target without affecting others.
* --dry-run walks the full pipeline but skips execSync; useful in
CI to validate manifest before bumping version.
Tested locally: clawhub auth detected, skillsmp/vercel CLIs not
installed (marked NOT READY and skipped cleanly in dry-run).
Follow-up work (tracked in TODOS.md later):
- Version-bump helper that reads openclaw/skills/*/SKILL.md frontmatter
and updates skills.json in lockstep.
- CI workflow that runs gstack-publish --dry-run on every PR and
gstack-publish on tags.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor: split preamble.ts into submodules (byte-identical output)
Splits scripts/resolvers/preamble.ts (841 lines, 18 generator functions +
composition root) into one file per generator under
scripts/resolvers/preamble/. Root preamble.ts becomes a thin composition
layer (~80 lines of imports + generatePreamble).
Before:
scripts/resolvers/preamble.ts 841 lines
After:
scripts/resolvers/preamble.ts 83 lines
scripts/resolvers/preamble/generate-preamble-bash.ts 97 lines
scripts/resolvers/preamble/generate-upgrade-check.ts 48 lines
scripts/resolvers/preamble/generate-lake-intro.ts 16 lines
scripts/resolvers/preamble/generate-telemetry-prompt.ts 37 lines
scripts/resolvers/preamble/generate-proactive-prompt.ts 25 lines
scripts/resolvers/preamble/generate-routing-injection.ts 49 lines
scripts/resolvers/preamble/generate-vendoring-deprecation.ts 36 lines
scripts/resolvers/preamble/generate-spawned-session-check.ts 11 lines
scripts/resolvers/preamble/generate-ask-user-format.ts 16 lines
scripts/resolvers/preamble/generate-completeness-section.ts 19 lines
scripts/resolvers/preamble/generate-repo-mode-section.ts 12 lines
scripts/resolvers/preamble/generate-test-failure-triage.ts 108 lines
scripts/resolvers/preamble/generate-search-before-building.ts 14 lines
scripts/resolvers/preamble/generate-completion-status.ts 161 lines
scripts/resolvers/preamble/generate-voice-directive.ts 60 lines
scripts/resolvers/preamble/generate-context-recovery.ts 51 lines
scripts/resolvers/preamble/generate-continuous-checkpoint.ts 48 lines
scripts/resolvers/preamble/generate-context-health.ts 31 lines
Byte-identity verification (the real gate per Codex correction):
- Before refactor: snapshotted 135 generated SKILL.md files via
`find -name SKILL.md -type f | grep -v /gstack/` across all hosts.
- After refactor: regenerated with `bun run gen:skill-docs --host all`
and re-snapshotted.
- `diff -r baseline after` returned zero differences and exit 0.
The `--host all --dry-run` gate passes too. No template or host behavior
changes — purely a code-organization refactor.
Test fix: audit-compliance.test.ts's telemetry check previously grepped
preamble.ts directly for `_TEL != "off"`. After the refactor that logic
lives in preamble/generate-preamble-bash.ts. Test now concatenates all
preamble submodule sources before asserting — tracks the semantic contract,
not the file layout. Doing the minimum rewrite preserves the test's intent
(conditional telemetry) without coupling it to file boundaries.
Why now: we were in-session with full context. Codex had downgraded this
from mandatory to optional, but the preamble had grown to 841 lines and
was getting harder to navigate. User asked "why not?" given the context
was hot. Shipping it as a clean bisectable commit while all the prior
preamble.ts changes are fresh reduces rebase pain later.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore: bump version and changelog (v0.19.0.0)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore: trim verbose preamble + coverage audit prose
Compress without removing behavior or voice. Three targeted cuts:
1. scripts/resolvers/testing.ts coverage diagram example: 40 lines → 14
lines. Two-column ASCII layout instead of stacked sections.
Preserves all required regression-guard phrases (processPayment,
refundPayment, billing.test.ts, checkout.e2e.ts, COVERAGE, QUALITY,
GAPS, Code paths, User flows, ASCII coverage diagram).
2. scripts/resolvers/preamble/generate-completion-status.ts Plan Status
Footer: was 35 lines with embedded markdown table example, now 7
lines that describe the table inline. The footer fires only at
ExitPlanMode time — Claude can construct the placeholder table from
the inline description without copying a literal example.
3. Same file's Plan Mode Safe Operations + Skill Invocation During Plan
Mode sections compressed from ~25 lines combined to ~12. Preserves
all required test phrases (precedence over generic plan mode behavior,
Do not continue the workflow, cancel the skill or leave plan mode,
PLAN MODE EXCEPTION).
NOT touched:
- Voice directive (Garry's voice — protected per CLAUDE.md)
- Office-hours Phase 6 Handoff (Garry's voice + YC pitch)
- Test bootstrap, review army, plan completion (carefully tuned behavior)
Token savings (per skill, system-wide):
ship/SKILL.md 35474 → 34992 tokens (-482)
plan-ceo-review 29436 → 28940 (-496)
office-hours 26700 → 26204 (-496)
Still over the 25K ceiling. Bigger reduction requires restructure
(move large resolvers to externally-referenced docs, split /ship into
ship-quick + ship-full, or refactor the coverage audit + review army
into shorter prose). That's a follow-up — added to TODOS.
Tests: 420/420 pass on gen-skill-docs.test.ts + host-config.test.ts.
Goldens regenerated for claude/codex/factory ship.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(ci): install Node.js from official tarball instead of NodeSource apt setup
The CI Dockerfile's Node install was failing on ubicloud runners. NodeSource's
setup_22.x script runs two internal apt operations that both depend on
archive.ubuntu.com + security.ubuntu.com being reachable:
1. apt-get update (to refresh package lists)
2. apt-get install gnupg (as a prerequisite for its gpg keyring)
Ubicloud's CI runners frequently can't reach those mirrors — last build hit
~2min of connection timeouts to every security.ubuntu.com IP (185.125.190.82,
91.189.91.83, 91.189.92.24, etc.) plus archive.ubuntu.com mirrors. Compounding
this: on Ubuntu 24.04 (noble) "gnupg" was renamed to "gpg" and "gpgconf".
NodeSource's setup script still looks for "gnupg", so even when apt works,
it fails with "Package 'gnupg' has no installation candidate." The subsequent
apt-get install nodejs then fails because the NodeSource repo was never added.
Fix: drop NodeSource entirely. Download Node.js v22.20.0 from nodejs.org as a
tarball, extract to /usr/local. One host, no apt, no script, no keyring.
Before:
RUN curl -fsSL https://deb.nodesource.com/setup_22.x | bash - \
&& apt-get install -y --no-install-recommends nodejs ...
After:
ENV NODE_VERSION=22.20.0
RUN curl -fsSL "https://nodejs.org/dist/v${NODE_VERSION}/node-v${NODE_VERSION}-linux-x64.tar.xz" -o /tmp/node.tar.xz \
&& tar -xJ -C /usr/local --strip-components=1 --no-same-owner -f /tmp/node.tar.xz \
&& rm -f /tmp/node.tar.xz \
&& node --version && npm --version
Same installed path (/usr/local/bin/node and npm). Pinned version for
reproducibility. Version is bump-visible in the Dockerfile now.
Does not address the separate apt flakiness that affects the GitHub CLI
install (line 17) or `npx playwright install-deps chromium` (line 33) —
those use apt too. If those fail on a future build we can address then.
Failing job: build-image (71777913820)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* chore: raise skill token ceiling warning from 25K to 40K
The 25K ceiling predated flagship models with 200K-1M windows and assumed
every skill prompt dominates context cost. Modern reality: prompt caching
amortizes the skill load across invocations, and three carefully-tuned
skills (ship, plan-ceo-review, office-hours) legitimately pack 25-35K
tokens of behavior that can't be cut without degrading quality or removing
protected content (Garry's voice, YC pitch, specialist review instructions).
We made the safe prose cuts earlier (coverage diagram, plan status footer,
plan mode operations). The remaining gap is structural — real compression
would require splitting /ship into ship-quick vs ship-full, externalizing
large resolvers to reference docs, or removing detailed skill behavior.
Each is 1-2 days of work. The cost of the warning firing is zero (it's
a warning, not an error). The cost of hitting it is ~15¢ per invocation
at worst, amortized further by prompt caching.
Raising to 40K catches what it's supposed to catch — a runaway 10K+ token
growth in a single release — without crying wolf on legitimately big
skills. Reference doc in CLAUDE.md updated to reflect the new philosophy:
when you hit 40K, ask WHAT grew, don't blindly compress tuned prose.
scripts/gen-skill-docs.ts: TOKEN_CEILING_BYTES 100_000 → 160_000.
CLAUDE.md: document the "watch for feature bloat, not force compression"
intent of the ceiling.
Verification: `bun run gen:skill-docs --host all` shows zero TOKEN
CEILING warnings under the new 40K threshold.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(ci): install xz-utils so Node tarball extraction works
The direct-tarball Node install (switched from NodeSource apt in the last
CI fix) failed with "xz: Cannot exec: No such file or directory" because
Ubuntu 24.04 base doesn't include xz-utils. Node ships .tar.xz by default,
and `tar -xJ` shells out to xz, which was missing.
Add xz-utils to the base apt install alongside git/curl/unzip/etc.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(benchmark): pass --skip-git-repo-check to codex adapter
The gpt provider adapter spawns `codex exec -C <workdir>` with arbitrary
working directories (benchmark temp dirs, non-git paths). Without
`--skip-git-repo-check`, codex refuses to run and returns "Not inside a
trusted directory" — surfaced as a generic error.code='unknown' that
looks like an API failure.
Benchmarks don't care about codex's git-repo trust model; we just want
the prompt executed. Surfaced by the new provider live E2E test on a
temp workdir.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(benchmark): add --dry-run flag to gstack-model-benchmark
Matches gstack-publish --dry-run semantics. Validates the provider list,
resolves per-adapter auth, echoes the resolved flag values, and exits
without invoking any provider CLI. Zero-cost pre-flight for CI pipelines
and for catching auth drift before starting a paid benchmark run.
Output shape:
== gstack-model-benchmark --dry-run ==
prompt: <truncated>
providers: claude, gpt, gemini
workdir: /tmp/...
timeout_ms: 300000
output: table
judge: off
Adapter availability:
claude: OK
gpt: NOT READY — <reason>
gemini: NOT READY — <reason>
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* test: lite E2E coverage for benchmark, taste engine, publish
Fills real coverage gaps in v0.19.0.0 primitives. 44 new deterministic
tests (gate tier, ~3s) + 8 live-API tests (periodic tier).
New gate-tier test files (free, <3s total):
- test/taste-engine.test.ts — 24 tests against gstack-taste-update:
schema shape, Laplace-smoothed confidence, 5%/week decay clamped at 0,
multi-dimension extraction, case-insensitive matching, session cap,
legacy profile migration with session truncation, taste-drift conflict
warning, malformed-JSON recovery, missing-variant exit code.
- test/publish-dry-run.test.ts — 13 tests against gstack-publish --dry-run:
manifest parsing, missing/malformed JSON, per-skill validation errors
(missing source file / slug / version / marketplaces), slug filter,
unknown-skill exit, per-marketplace auth isolation (fake marketplaces
with always-pass / always-fail / missing-binary CLIs), and a sanity
check against the real repo manifest.
- test/benchmark-cli.test.ts — 11 tests against gstack-model-benchmark
--dry-run: provider default, unknown-provider WARN, empty list
fallback, flag passthrough (timeout/workdir/judge/output), long-prompt
truncation, prompt resolution (inline vs file vs positional), missing
prompt exit.
New periodic-tier test file (paid, gated EVALS=1):
- test/skill-e2e-benchmark-providers.test.ts — 8 tests hitting real
claude, codex, gemini CLIs with a trivial prompt (~$0.001/provider).
Verifies output parsing, token accounting, cost estimation, timeout
error.code semantics, Promise.allSettled parallel isolation.
Per-provider availability gate — unauthed providers skip cleanly.
This suite already caught one real bug (codex adapter missing
--skip-git-repo-check, fixed in
|