v1.57.5.0 feat: cross-session decision memory + gbrain dream-stage call graph#1910
Merged
Conversation
Reads `gbrain doctor` cycle_freshness to classify whether a source has completed a full cycle (completed/never/unknown). A fail naming this source -> never; a fail naming only other sources -> completed; an absent or unparseable check -> unknown, so an unrelated doctor failure never masks a real state. Gates the automatic call-graph build on --full. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…est outcome guard Adds a source-scoped `gbrain dream --source <id>` stage that builds this worktree's call graph (code-callers/code-callees). Runs lock-free after the sync lock releases so it never blocks sibling worktrees; a .dream-in-progress marker dedupes concurrent dreams. --full auto-runs it only when the cycle was never built; explicit --dream always forces; --no-dream opts out. The stage parses the cycle's own output and reports the truth, not a flat "built": a WARN when the schema pack can't extract code symbols, when the embed phase failed for a missing key, or when 0 edges resolved; OK with the resolved-edge count otherwise. gbrain exits 0 even when it skips on a held cycle lock (e.g. autopilot), so that case reports SKIP, not success. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
gbrain writes per-source staging and capability-check artifacts under .sources/ in the repo root. It's machine-local runtime state, not source. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…on gbrain>=0.41.38 sync-gbrain frames the --dream offer honestly: building a call graph requires a code-aware schema pack, and the dream stage reports a WARN when it can't. The verdict's Call graph row mirrors the dream stage's real outcome instead of assuming a completed cycle means edges exist. The ## GBrain Search Guidance block written into CLAUDE.md drops the old code-callers --source caveat: gbrain >=0.41.38.0 honors the .gbrain-source pin for code-callers/code-callees. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…in-gstack # Conflicts: # bin/gstack-gbrain-sync.ts # lib/gbrain-sources.ts
…atomic append + tolerant read) Single source of truth extracted for D2A: gstack-learnings-* and the upcoming gstack-decision-* bins share one injection-pattern list, one atomic single-line appender, and one tolerant reader. No more drift between stores. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
… (D2A) Replace the inline injection-pattern copy with the shared list. One audited write-path rejection across learnings + the upcoming decision store. Behavior unchanged (35/35 learnings tests green); learnings-search keeps its inline copy because a structural test pins its bash/bun shape. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ion) decide/supersede/redact events on lib/jsonl-store; active set is computed (no mutable status), dangling refs tolerated. Free-text is injection-checked and redact-scanned on write (HIGH secret -> reject). Scope filter (repo/branch/issue) for relevant resurfacing. File-only + reliable; gbrain not required. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…, supersede archives) writeSnapshot/readSnapshot/rebuildSnapshot give an O(active) bounded read for the session-start hot path (D1A). compact() rewrites the log to active, archives superseded decisions for history, and EXPUNGES redacted ones (dropped, never archived) so an accidentally-captured secret leaves the store for good. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…n-interactive) Two bins mirroring gstack-learnings-* (D3A). log writes decide/--supersede/--redact/ --compact events + refreshes the bounded snapshot + enqueues for cross-machine sync; search reads the O(active) snapshot, scope-filtered to current branch, newest-first, --all to include superseded, --json for machines. Empty store returns silently (no snapshot write on an empty read). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ge (Context Recovery) Context Recovery now shows recent scope-relevant active decisions (bounded read of decisions.active.json via gstack-decision-search) and instructs the agent to treat them as settled calls and to log durable decisions/reversals. Closes the Phase-1 capture->curate->resurface loop, reliable + file-only. Regen across all hosts folded in (squash-with-regen); parity 10/10, freshness green. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Context Recovery now emits the cross-session-decisions block, so ship's preamble (all hosts) changed. Golden baselines are hand-maintained copies (gen does not write them); refresh them from the fresh gen so golden-file regression passes. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…DE.md Adds a '## Cross-session decision memory' section: how to resurface (gstack-decision-search) and capture (gstack-decision-log) durable decisions, the supersede/redact/compact verbs, and a crisp durable-vs-trivial definition so the store stays signal. Reliable file-only path; gbrain not required. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ured points Wires the four skills that finalize real decisions to capture them in the cross-session decision store, from their STRUCTURED outputs (never free-text scraping): - ship: the version bump (level + why) at write time - plan-ceo-review: accepted scope + verdict (branch-scoped) - plan-eng-review: the architecture verdict + key call (branch-scoped) - spec: the filed issue's core approach (issue-scoped) All emits are non-interactive, schema-correct (content in decision/rationale, source=skill, confidence 1-10), and best-effort (|| true) so a decision-log failure never blocks the workflow. Includes regen across hosts + refreshed ship golden baselines. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds gstack-decision-search --semantic (with --query): appends a 'Related from memory' block from gbrain semantic search, scoped to the curated-memory source. Pure enhancement, reliability-first: a new lib/gstack-decision-semantic.ts is the ONLY decision module that touches gbrain and is imported lazily only on --semantic, so the reliable file path never loads gbrain code. Every path degrades to the reliable file results when gbrain is off, unconfigured, empty, or errors (never throws, 10s timeout). Built against the verified gbrain 0.42.x surface (text output [score] slug -- snippet, NOT JSON; curated-memory source resolved by worktree path, not a gstack-brain-<user> id). Deterministic-contract tests only: parser units, degrade-to-null when gbrain absent, and a fake-gbrain shim proving scope+search end-to-end. find-contradictions deferred (no verifiable CLI surface yet + curated memory not indexed). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
detectAutopilot treated a lock FILE as proof of life, so a crashed gbrain daemon left a stale lock that wedged every sync forever (observed: a dead pid refused --full indefinitely). Now read the holder pid (bare or JSON body) and check liveness via signal-0: ESRCH=dead → ignore the stale signal and keep checking; EPERM=alive (other user) → active. A stale lock never masks a live autopilot process. Pure decision function — does not delete the file; the caller may clean it. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…in-use # Conflicts: # plan-ceo-review/SKILL.md # plan-ceo-review/SKILL.md.tmpl # plan-eng-review/SKILL.md # plan-eng-review/SKILL.md.tmpl
…keys
Pre-existing on main (v1.56.x): the two section-loading E2E tests used
human-label testNames ('/ship section-loading') that don't match their slug
keys ('ship-section-loading') in E2E_TOUCHFILES/E2E_TIERS. Every other E2E test
uses the slug as its testName, and the TOUCHFILES completeness gate requires
testName to be a registered key — so the gate was red. Align both testNames to
their slug keys (also fixes tier lookup for these two periodic tests).
Verified failing on a clean origin/main checkout before the fix.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Addresses the pre-landing review findings (all INFORMATIONAL, no criticals): - security: datamark resurfaced decision text at the render boundary (lib/gstack-decision.ts datamark() — neutralizes code fences, --- banners, <|role|>/</system> markers, control chars, newlines). Applied in gstack-decision-search human output so stored text can't masquerade as instructions in Context Recovery (codex hardening #3 / AC #7). --json stays raw. - DRY: extract resolveSlug/gitBranch/flagValue to lib/bin-context.ts; both decision bins use it instead of duplicating the helpers. - compact(): batch the archive append (one write, not N) and shrink the mid-compact crash window; simplify the opaque branch/issue ternary. - coverage: learnings-log injection rejection (D2A wiring), search --recent/ --scope + NaN-safe --recent, datamark-applied, unparseable lock body, compact-empty, corrupt-snapshot degrade. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adversarial review (Claude subagent) found a CRITICAL the specialist pass missed: - F1 (CRITICAL): 'Human:'/'Assistant:' turn-prefixes bypassed BOTH the write-time denylist AND datamark(), landing verbatim in agent context inside the trusted ACTIVE DECISIONS fence. Add 'human:' (+ 'disregard previous', 'from now on') to the shared denylist, and have datamark() neutralize Human:/Assistant:/System:/User: turn-prefixes (ZWSP) at the render boundary. - F2: datamark() only stripped ASCII C0; extend to Unicode line terminators (U+0085/2028/2029) and U+007F so 'strip newlines' actually holds. - F3: validateDecide blocked only HIGH secrets; MEDIUM-tier PII (e.g. SSN) persisted silently and synced cross-machine. The store is non-interactive (no confirm path), so fail closed on MEDIUM too. - F4: compact() was a lock-free read-modify-rewrite that could clobber a concurrent append (lost decision). Add an O_EXCL compact lock + a pre-rename size recheck that aborts untouched (skipped=true) if an append landed; caller re-runs. - F7: filterByScope unknown/garbage scope fell through to 'return true' (leaked into every context); fail conservative (false). F5 (pid reuse) and F6 (pgrep over-match) are intentionally left as-is: both fail SAFE (over-refuse sync); making them precise would introduce a fail-DANGEROUS path (allowing sync during a real autopilot). True disambiguation needs gbrain to stamp the lock with a start-time, which gstack doesn't own. F8 (compact moves history to archive) is by design. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Codex adversarial review found a HIGH the Claude pass missed plus 3 mediums: - C1 (HIGH): gstack-decision-search --all returned every decide and IGNORED redact events, so a redacted secret still resurfaced via --all until compact ran. --all now excludes redacted (redact = expunge from every read path), still showing superseded history. - C-med: semantic (external gbrain) slug/snippet were printed raw — datamark them too so a gbrain hit can't spoof role markers / fences into agent context. - C4: semanticRecall fell back to an UNSCOPED gbrain search when no curated-memory source resolved, pulling code/doc corpora mislabeled as 'related decisions'. Now returns null (degrade) when there's no worktree-backed memory source. - C5: validateDecide scanned only decision/rationale/alternatives; branch and issue are stored + surfaced (raw via --json), so include them in the injection+secret scan. C2 (snapshot staleness) / C3 (compact TOCTOU residual): accepted for a single-user store — atomic appends never lose the event, rebuilds self-heal, and the compact size-recheck leaves only a sub-ms window; full append-locking would break the lock-free append design. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
Merging to
After your PR is submitted to the merge queue, this comment will be automatically updated with its status. If the PR fails, failure details will also be posted here |
Resolves VERSION/CHANGELOG/package.json (branch keeps its higher v1.57.5.0; my CHANGELOG entry on top, main's v1.57.2.0 entry below). Regenerated SKILL.md across hosts from the merged templates. Bumped maxSizeRatio to 1.07 for investigate, cso, and design-consultation: the cross-cutting preamble growth (v1.57.2.0 AUQ-failure prose fallback + the decision-memory nudge) lands them just over the strict 1.05. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Resolves VERSION/CHANGELOG/package.json (branch keeps v1.57.5.0; my entry on top, main's v1.57.3.0 entry below). Regenerated SKILL.md across hosts from merged templates. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
E2E Evals: ❌ FAIL53/59 tests passed | $11.56 total cost | 12 parallel runners
12x ubicloud-standard-8 (Docker: pre-baked toolchain + deps) | wall clock ≈ slowest suite Failures
|
|
This is a master class in institutional memory architecture — the "capture → curate → resurface" loop you've built solves the hardest problem in cross-session agent systems: how do decisions persist without becoming noise? Why this matters beyond gstack:
Two design notes that stand out:
The adversarial review rigor is rare:
Comparison to the ecosystem: Most agent systems treat memory as append-only logs with no curation or explicit invalidation. The result is context pollution — old decisions resurface and confuse new sessions. Your Similar event-sourced memory patterns in SwarmAI. Discussion: T-MEM |
Resolves VERSION/CHANGELOG/package.json (branch keeps v1.57.5.0; my entry on top, main's v1.57.4.0 Boil-the-Ocean rename entry below). Regenerated SKILL.md across hosts (picks up the Boil-the-Lake to Boil-the-Ocean rename). Bumped plan-eng-review maxSizeRatio to 1.06 for the cumulative cross-cutting preamble growth. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This was referenced Jun 13, 2026
stevenbarragan
added a commit
to stevenbarragan/gstack
that referenced
this pull request
Jun 15, 2026
…romium-launch-headed * origin/main: (35 commits) v1.58.1.0 feat: hermetic local E2E + Conductor prose AskUserQuestion (garrytan#2004) v1.58.0.0 feat: diagram + multi-format document engine (mermaid, excalidraw, single-file HTML, DOCX) (garrytan#1990) v1.57.10.0 feat: Codex review default-on across review/ship/plan/docs (garrytan#1966) v1.57.9.0 feat: source-clean gbrain render (dev-setup --out-dir + machine-wide gbrain-refresh) (garrytan#1951) v1.57.8.0 feat: browse js/eval --out render-to-file (canonical Chromium for offline rendering) (garrytan#1929) v1.57.7.0 feat: GSTACK REVIEW REPORT always declares unresolved decisions (garrytan#1916) v1.57.6.0 fix wave: 8 community bugs (4 security guards failing open) (garrytan#1911) v1.57.5.0 feat: cross-session decision memory + gbrain dream-stage call graph (garrytan#1910) v1.57.4.0 refactor(ethos): rename Boil the Lake principle to Boil the Ocean (garrytan#1912) v1.57.3.0 fix(ship): always-loaded PR-title-version rule + fork-PR title-sync backstop (garrytan#1909) v1.57.2.0 feat: AskUserQuestion prose fallback when the tool fails at runtime (garrytan#1908) v1.57.0.0 feat: carve-guard system + carve cso/document-release/design-consultation (garrytan#1907) v1.56.1.0 fix(sync): staging-dir ownership guard + resume-correctness fixes (garrytan#1802) (garrytan#1856) v1.56.0.0 Token-reduction Phase B + AUQ paranoid safety net (garrytan#1849) v1.55.1.0 fix: telemetry consent accuracy + gstack-slug cache sanitization (garrytan#1848) v1.55.0.0 fix wave: gbrain data-loss guards + browser crash-loop + 6 more (garrytan#1808) fix(setup): add missing gen:skill-docs:user script (garrytan#1807) v1.54.0.0 feat: carve /ship into skeleton + on-demand sections (-59% always-loaded) (garrytan#1806) v1.53.1.0 fix: non-interactive-safe plan-tune hook install (flags + smart defaults) (garrytan#1805) v1.53.0.0 feat: smarter redaction — PII/secrets/legal guard across /spec, /ship, /cso, /document-* (garrytan#1797) ...
garrytan
added a commit
that referenced
this pull request
Jun 21, 2026
… gate (#2077) * fix(gbrain): stop forcing GBRAIN_PREPARE on transaction-mode poolers (#1965) buildGbrainEnv auto-set GBRAIN_PREPARE=true whenever DATABASE_URL targeted port 6543, and the /sync-gbrain capability check exported it for the rest of the skill run. Both had the semantics inverted: gbrain auto-disables prepared statements on transaction-mode poolers because they break every write there ("prepared statement does not exist"); GBRAIN_PREPARE=true is gbrain's documented override for SESSION-mode poolers on 6543, not a requirement for transaction mode. The #1435 search symptom the auto-set worked around was fixed gbrain-side. Remove both force-sets. A caller-set GBRAIN_PREPARE (either value) still passes through untouched, preserving the session-mode-on-6543 escape hatch. isTransactionModePooler stays exported. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(gbrain): classify probe timeout as its own status; sync proceeds instead of skipping (#1964) The 5s engine probe misclassified healthy-but-slow engines (cold Supabase pooler connections measured at 6.9-10.7s) as broken-config, so /sync-gbrain silently skipped code+memory and told the user their config was malformed. - New "timeout" status: probe killed at the deadline with no recognized stderr pattern. Default deadline is now 15s, overridable via GSTACK_GBRAIN_PROBE_TIMEOUT_MS (tests set 300ms against a fake that sleeps 2s). - Sync stages PROCEED on timeout with a stderr warning naming the env knob; a genuinely-dead engine surfaces its real error at the first operation instead of a false config diagnosis. - Consistency everywhere "ok" gated behavior: gstack-gbrain-detect --is-ok exits 0 on timeout, and gen-skill-docs' detection gate accepts it, so a slow engine no longer silently suppresses brain-aware features. - Status cache: key now includes the effective probe timeout (raising it invalidates a cached timeout) and GBRAIN_HOME; config detection honors GBRAIN_HOME so relocated-home users stop being misclassified as missing-config. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(bins): cygpath-normalize SCRIPT_DIR for bun imports; surface learnings-log errors (#1950) Under Windows git-bash, pwd yields a POSIX path (/c/Users/...) that Bun on Windows cannot resolve as an ES module specifier. gstack-learnings-log interpolates SCRIPT_DIR into a bun -e import, so every invocation died with "Cannot find module" — and 2>/dev/null swallowed the error, silently dropping every AI-logged learning for Windows users. - 3-line cygpath -m guard in gstack-learnings-log and gstack-question-log (which gains the same import shape in the next commit). Matches the duplicated IS_WINDOWS convention in setup; no shared shell lib exists. - learnings-log adopts question-log's set +e / TMPERR capture pattern wholesale: validation errors now print to stderr. The old `if [ $? -ne 0 ]` check was dead code under set -euo pipefail — the script exited at the failing assignment before reaching it. - New test/bin-windows-bun-import-paths.test.ts: static invariant (any bash bin interpolating $SCRIPT_DIR into a bun -e import must carry the guard) + behavioral end-to-end run invoked via `bash <bin>` — added to the windows-free-tests workflow list so the conversion is proven on the only platform where the bug exists. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(question-log): dedupe INJECTION_PATTERNS via lib/jsonl-store (#1934) bin/gstack-question-log carried a local copy of the injection-pattern list, so pattern fixes to lib/jsonl-store.ts never propagated — including the /override[:\s]/i false-positive fix arriving via community PR #1940. Import the shared hasInjection instead (enabled by the previous commit's cygpath guard). question-log also gets the lib's stricter superset (human:, disregard, from-now-on, approve-all patterns). Tests pin the contract in a #1940-order-independent way: an "Override: ignore all previous instructions" header is rejected, "prose overrides the deterministic table" is accepted, and a static invariant keeps local INJECTION_PATTERNS duplicates out of the bin. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(security): community-pulse + both dashboards never report fake zeros (#1947) The security-signaling surface failed open at three layers — every failure mode read as a reassuring "0 attacks" / "0 installs": - community-pulse edge function: supabase-js returns {data,error} without throwing, and all five queries discarded `error` — a DB outage produced real-looking zeros via the SUCCESS path, and the catch (also returning zeros with HTTP 200) was unreachable for query failures. Every query now destructures and throws; the catch serves the stale cache (marked "stale": true) when one exists, else 503 {"error":"pulse_unavailable"}. Success responses carry "status":"ok" so clients can distinguish authoritative data from legacy backends. NOTE: the edge function deploys out-of-band (supabase functions deploy community-pulse). - gstack-security-dashboard: captures the HTTP status; non-200 / network failure / error body / missing section → "unknown — backend error"; jq missing → "unknown — install jq" (the lossy grep fallback broke on nested arrays and under-reported attacks as zero — removed); a 200 without the new marker shows figures with an "unverified (legacy backend)" note. Also fixes a latent display bug: the TOTAL grep matched the digit 7 inside "attacks_last_7_days" and misreported every count. - gstack-community-dashboard: same class — curl || echo "{}" plus grep || echo "0" printed "Weekly active installs: 0" on any failure. Now "unknown — backend error (HTTP N)". test/security-dashboard-fallback.test.ts pins the matrix (200+marker, 200-legacy, 503, network failure) x (jq present, jq absent) for both bins: "unknown" states never render as 0. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(telemetry): redact error_message spans before they leave the machine (#1947) error_message was uploaded with only quote/newline escaping — stack traces and failed-API errors can embed credentials, private paths, and hostnames, and the sync path strips only _repo_slug/_branch. New lib/redact-engine.ts export redactFindingSpans(): replaces EVERY finding's span with <REDACTED-{id}> regardless of tier (applyRedactions is the interactive PII-only path and exits nonzero on credential findings, so it can't serve machine egress). Returns null when a span can't be located — callers drop the whole payload rather than risk a leak. gstack-telemetry-log pipes error_message through it at LOG time, so the local JSONL at rest is clean too; surrounding text survives for crash triage. FAIL CLOSED: bun missing, engine error, or non-JSON-string output all null the field. Tests pin: embedded ghp_ token → <REDACTED-github.pat> with context intact; redactor unavailable → null; raw bytes on disk never contain the token. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(redact): prepush guard fails closed on git failure; /ship owns hook install (#1946) Two gaps closed: 1. Fail closed. The git() helper returned "" on ANY non-zero exit or maxBuffer overflow (status null), addedLinesFor produced an empty string, and the push sailed through unscanned — fail-open on exactly the oversized-diff case where a large secret-bearing blob is most likely. The diff call now uses a strict variant that throws; main blocks with a clear message naming the GSTACK_REDACT_PREPUSH=skip escape valve. Probe calls (symbolic-ref, rev-parse, merge-base) keep the permissive helper — their failures are normal control flow. 2. Install path. The hook was installed by nothing ("opt-in, installed by nothing" was the issue's words). ./setup runs in the gstack checkout — the wrong repo for a per-project hook — so it gets a one-line hint only. /ship owns per-repo install: config redact_prepush_hook=true + hook missing → silent install (consent already given); config unset + no ~/.gstack/.redact-prepush-prompted marker → one-time machine-wide AskUserQuestion offer, answer persisted. ship/SKILL.md regenerated in this same commit (check-freshness bisect discipline). Tests: unscannable diff (bogus SHAs) → exit 1 + valve named; empty-but- successful diff → exit 0; static asserts pin setup as hint-only and the ship template as the installer surface. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(redact): six new credential patterns — GitLab, HuggingFace, npm, DigitalOcean, Bearer, GCP SA (#1946) Coverage gaps from the #1946 security review, including token types for tooling gstack itself drives (glab): HIGH (block): gitlab.token (glpat-/glptt-/gldt-), huggingface.token (hf_), npm.token (npm_), digitalocean.token (dop_v1_), gcp.service_account (the JSON-escaped "private_key" form that dodges pem.private_key's literal-block match when minified, confirmed by "private_key_id" proximity). MEDIUM (warn): auth.bearer — the most FP-prone shape in the set (docs are full of "Authorization: Bearer <token>"), so it requires header-context proximity and the same entropy>=3.0 + placeholder validator recipe as env.kv. "Bearer YOUR_TOKEN_HERE" never fires; calibration over coverage, per the cries-wolf principle. All shapes are linear-time; test/redact-pattern-lint.test.ts covers them automatically. Engine tests add positive + placeholder-negative cases per pattern. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * test: coverage-audit additions for the fix wave Ship Step 7 gap-fill (all passing, 248 tests across the touched suites): memory + dream stage probe-timeout proceeds, gbrain-detect override paths, stale-flag passthrough, 200-body-missing-.security fail-closed case, telemetry redaction edges, and credential-pattern edge cases. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: pre-landing review fixes Review army findings (1 critical, auto-fixed with regression tests): - CRITICAL (security specialist, verified live): redactFindingSpans spliced only the regex capture span, and pem.private_key / gcp.service_account capture just the BEGIN-header — the key body survived "redaction" and shipped via telemetry. Marker-only patterns now drop the whole payload (null, fail closed). Overlapping spans (Bearer+JWT on the same bytes) are coalesced before splicing so stale offsets can't leave partial secret bytes behind. - gitStrict: drop the dead `|| r.status === null` disjunct (null !== 0 already covers it); add the signal-kill/null-status regression test the docstring promised. - security-dashboard human mode flags stale snapshots ("figures may be out of date") instead of presenting frozen counts as current. - community-dashboard marker check uses jq when available — the grep-only variant misclassified whitespaced/reserialized bodies as legacy. - telemetry fail-closed test now shadows bun with a failing stub (deterministic on any host layout); stale "five status cases" describe title renamed. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: adversarial review fixes (Claude + Codex cross-model passes) Both adversarial passes ran against the wave; every FIXABLE finding landed with a regression test: - probeTimeoutMs clamps to >=1ms: a fractional override floored to 0, and execFileSync treats timeout:0 as NO timeout — the probe that exists to bound hangs could hang forever (found by both models independently). - /ship silent hook install now requires the hooks dir to live inside .git: with core.hooksPath (husky's COMMITTED .husky/), the chaining installer would have renamed the team's committed pre-push and written a machine-local wrapper into the working tree (found by both models). - gstack-config gbrain-refresh accepts the "timeout" status — the last consumer still gating on literal "ok" (Codex); gstack-gbrain-detect's config-derived fields honor GBRAIN_HOME so the detection JSON can't report status ok alongside config_exists false (Codex). - prepush: a remote sha absent locally (shallow clone / stale fetch) falls back to the merge-base/empty-tree range — scans MORE, never blocks a legitimate push into training users toward --no-verify. - dashboards: curl's own 000 no longer doubles to "HTTP 000000"; the community dashboard flags stale snapshots like the security one; array sections parse via jq (the sed/grep loops truncated at the first ']'); the no-jq marker grep tolerates whitespace. - telemetry: multi-line redactor output nulls the field instead of corrupting the JSONL record; setup's hint fires only when the config key is genuinely unset (an explicit false is a recorded decline); the /ship prompt marker honors GSTACK_HOME. Kept as designed (cross-model tension noted): Bearer stays MEDIUM in the prepush gate — a HIGH Bearer would block every docs example; the entropy validator can't eliminate that FP class, and MEDIUM warns visibly. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * chore: bump version and changelog (v1.57.11.0) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs: P1 TODO — eval harness live progress + incremental persistence Root-caused during this ship: a killed eval run was indistinguishable from a healthy one for hours (per-file output buffering across mega test files, no incremental eval-store writes, no honest liveness signal). Full context and starting points in the entry. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * test: fix operational-learning E2E fixture — copy lib/jsonl-store.ts Pre-existing breakage, proven on main: gstack-learnings-log has imported lib/jsonl-store.ts (shared injection patterns) since v1.57.5.0 / #1910, but the fixture copies only the bin scripts — the bin exits 1 before writing anything, on main silently (stderr swallowed) and on this branch loudly (the #1950 error-surfacing made the four-day-old failure visible). A real install always ships bin/ and lib/ together; the fixture now does too. Verified: the fixture-shaped invocation writes the learning (exit 0) with lib present, exits 1 on both main and this branch without it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(ios-qa): isolate E2E tests under --concurrent (3 real races) The ios-qa E2E file failed intermittently under `bun test --concurrent` (the eval harness default). Three distinct shared-state races, all fixed: 1. Shared pidfile: a module-level `workDir` reassigned in beforeEach was clobbered by parallel tests, so concurrent daemons collided on the same pidfile and the loser returned `already_running`. Each test now gets its own dir via makeWorkDir(). 2. process.env path globals: tests set GSTACK_IOS_AUDIT_PATH / _ATTEMPTS_PATH / _ALLOWLIST_PATH on the shared process env; concurrent tests stomped each other's audit/attempts destinations. Threaded auditPath/attemptsPath/allowlistPath through DaemonOptions (and mintForCaller) as explicit args — env is no longer load-bearing. 3. afterEach cleanup race: the per-test cleanup drained a shared dir array, so the first test to finish deleted still-running tests' workDirs mid-assertion. Moved to afterAll (cleans once, after all settle). Verified: 5/5 clean full-suite runs at --max-concurrency 15 (was intermittent); daemon unit suite 91/91; daemon source compiles. The paths default to the env-derived locations when options are omitted, so the production CLI path is unchanged. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * test(pty): pin spawned claude to EVALS model chain (default claude-sonnet-4-6) launchClaudePty spawned the interactive `claude` TUI with no --model flag, so the child inherited the operator's ~/.claude/settings.json model. On a slow-thinking model that meant 5+ min of extended thinking on empty plan-mode context, timing out the plan-mode smoke tests regardless of contention. Pin the model via opts.model ?? EVALS_MODEL ?? 'claude-sonnet-4-6' — byte-identical to session-runner.ts:144, so PTY and `claude -p` evals always agree. Pushed before extraArgs (last flag wins, so a per-test --model still overrides). Placement leaves the spawn region byte-stable for a clean merge with the in-flight hermetic-env branch. Plumbed model through the three plan-skill wrappers. Static-grep tripwires guard the pin, its fallback chain, the before-extraArgs ordering, and all three wrapper forwards. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(pty): detect markdown bold-bullet prose AUQs (fixes office-hours smoke) office-hours auto-mode renders its mode question as `- **Building a startup**` markdown bullets (office-hours/SKILL.md.tmpl:102) with no letter/number marker. isProseAUQVisible only matched `A)`-style lettered or `1.`-style numbered options, so the question went undetected: the model surfaced it at ~2m19s (well under the 300s budget) but the harness kept scoring the run "working" off the spinner glyphs and timed out — a false timeout on a question that was already on screen. Add Pattern 3: when an interrogative line ('?') is present AND 3+ bold-bullet markers (`- **`) appear in the 4KB tail, classify as a prose AUQ. Bold is the discriminator vs incidental prose bullets; the line anchor is dropped (stripAnsi can collapse option lines) and the existing `❯ 1.` cursor gate still defers to a live native list. Wires through the existing classifyVisible 'asked' path and the timeout high-water-mark, so office-hours now classifies 'asked' instead of 'timeout'. Five unit cases: the office-hours render passes; no-'?', <3-bullet, plain-bullet, and native-cursor cases stay false. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(pty): detect stripAnsi-collapsed prose AUQs + judge spinner-precedence The plan-eng/plan-design plan-mode + finding-floor smokes timed out even when the skill HAD rendered a complete prose AskUserQuestion and was waiting: the PTY strips cursor-positioning escapes, collapsing the option newlines/spaces so "A) ..." arrives as "A(recommended)" / "-B:" and "Reply with A, B, or C" as "ReplywithA,B,orC". Every line-anchored detector (Patterns 1-3) returns false on those bytes, so proseAUQEverObserved never latched and the run timed out on a question that was already on screen. Add Pattern 4/5: a two-signal collapsed-form detector — a reply/recommendation marker (space-insensitive "reply with [A-D]", "Recommendation:", or "(recommended)") AND 2+ distinct A-D letters each punctuated by ) : or (. The conjunction is what separates a real AUQ from incidental report prose; verified true on the verbatim failing-run buffers where Patterns 1-3 return false. Also fix the Haiku judge spinner bias: of 614 verdicts, 569 were 'working' and 95 of those noted a question was visible — Claude Code keeps the spinner animating at an idle prose decision, so the judge coin-flipped. Add a precedence override: when an option list AND a Recommendation/Reply instruction are both visible, classify WAITING even with spinner glyphs. Kept the strict dual-signal gate (never option-list-alone) so auto-decide-preserved doesn't flip. 5 unit tests pin the two-signal contract (2 true on real collapsed bytes, 3 false guards). 90 -> 95 pass. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(plan-review): ask-first scope gate for plan-eng + plan-design review On an empty/cold invocation, plan-eng-review and plan-design-review would dive straight into repo exploration (plan-eng) or a 7-pass mockup+audit (plan-design) and only ask the user much later, if at all. plan-ceo-review already asks first via an unconditional Step-0 gate and behaves well; these two did not. Add a hard-STOP scope gate as the FIRST operational instruction in each skill (above the design-doc check / pre-review audit / mockup defaults it explicitly overrides): the first tool call must be AskUserQuestion confirming the review target, before any git/Read/Grep/Glob/Bash or mockup generation. Under --disallowedTools the options render as plain column-0 lettered prose with a Recommendation + "Reply with A, B, or C" line so the answer is detectable. This is correct cold-start UX (confirm what to review before grinding a full review on nothing) and it is the product half of the plan-mode smoke fix; the harness collapsed-form detector is the deterministic half that catches the ask however it renders. Templates + regenerated SKILL.md (default variant). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test(tiers): reclassify stochastic plan-eng/plan-design ask-first smokes as periodic plan-eng-review and plan-design-review run a long explore/audit before their first AskUserQuestion, so whether the plan-mode + finding-floor smokes reach a terminal outcome within the 300s/600s budget depends on stochastic ask-first compliance (measured ~50-67%/run even with the hardened gate). Per the "non-deterministic -> periodic" tiering rule, move the four affected smokes (plan-eng/plan-design review-plan-mode + finding-floor) to periodic. The deterministic harness fix (collapsed-form detector + judge precedence) and the ask-first gate lift these from always-failing to mostly-passing and are the real product+harness improvements; periodic monitoring tracks the rate weekly without blocking PRs on an LLM coin-flip. plan-ceo/plan-devex ask-first reliably and stay gate-tier. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * ci(evals): gate the deterministic PTY plan-mode smokes in CI The real-PTY plan-mode smokes never ran in CI — the gate was local-only. Add an e2e-pty-plan-smoke matrix suite running the two deterministically-reliable ones (office-hours-auto-mode, plan-mode-no-op) so a regression there blocks PRs. The stochastic plan-eng/plan-design ask-first smokes stay periodic (touchfiles E2E_TIERS) and are not CI-gated. A fresh CI container has no ~/.claude.json, so the spawned interactive `claude` would wedge on the onboarding + API-key-approval dialog. Add a scoped seed step (hasCompletedOnboarding + key approval, its own ANTHROPIC_API_KEY env) before the run — mirrors what the hermetic E2E child env seeds. Per-suite timeout override (35 min) via matrix.suite.timeout so the PTY suite has headroom for --retry 2 without bumping the other 12 suites. Report runner count 12 -> 13. Validate via workflow_dispatch before relying on the gate (PTY-in-CI is new). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * ci(evals): install gstack skill registry for the PTY smoke suite The first dry-run of e2e-pty-plan-smoke failed: the spawned interactive `claude` printed "Unknown command: /plan-ceo-review". .claude/skills is gitignored, so a fresh CI checkout has no gstack skill registry and the TUI can't resolve /office-hours or /plan-ceo-review. Add a Register step (scoped to the suite, after Seed, before Run) that mirrors setup's --no-prefix user-scoped registry minimally: $HOME/.claude/skills/gstack -> repo (resolves the preambles' absolute ~/.claude/skills/gstack/bin/* and <skill>/sections/* paths) + per-skill SKILL.md/sections symlinks for the two skills these tests invoke. HOME is /github/home in this container and the runner adds no HOME/CLAUDE_CONFIG_DIR override (no hermetic mode), so $HOME is the right anchor — the Seed step already proved claude reads it. No ./setup (binary build + Chromium + fonts + /dev/tty prompt); SKILL.md + bin/ + sections/ are committed. Self-validating: fails the step loudly on a dangling symlink or missing `name:` frontmatter, so a moved target surfaces here instead of as a silent 35-min "Unknown command" timeout. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore: bump version and changelog (v1.58.4.0) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Cross-session decision memory (the headline). An institutional memory loop so durable decisions and their rationale survive across sessions: capture → curate → resurface.
decide/supersede/redact) at~/.gstack/projects/<slug>/decisions.jsonl; "active" is computed, dangling-ref tolerant. Bounded active snapshot keeps session-start reads O(active), with explicit--compact.gstack-decision-log(--supersede/--redact/--compact) andgstack-decision-search(--scope/--recent/--query/--all/--json/--semantic).## Cross-session decision memorysection added to CLAUDE.md./plan-ceo-review,/plan-eng-review,/spec,/shipauto-capture their structured decisions.--semanticrecall, lazily loaded; degrades to the reliable file results when gbrain is off/empty.lib/jsonl-store.ts(injection-reject + atomic append + tolerant read; learnings-log refactored onto it) andlib/bin-context.ts.---banners,<|role|>/</system>, chat turn-prefixes, Unicode line terminators), HIGH and MEDIUM secrets blocked on write,redactexpunges from every read path (incl.--all), compact is lock + size-recheck guarded.gbrain dream-stage + reliability.
/sync-gbrain--dreambuilds the symbol cross-reference call graph behind a lock-free gate with an honest WARN-not-false-success guard;cycleCompleted()cycle-state probe./sync-gbrain; ignore gbrain.sources/staging dir.Test Coverage
All new code paths have test coverage. 117 tests across the decision store + gbrain stages (
jsonl-store,gstack-decision,gstack-decision-bins,gstack-decision-semantic,gbrain-dream-stage,gbrain-cycle-completed,gbrain-guards). Fullbun testpasses with 0 failures;gen:skill-docsclean (0 stale), parity 10/10, ship goldens refreshed.Pre-Landing Review
4 findings, all INFORMATIONAL (no criticals), all auto-fixed: datamark resurfaced text, DRY-extract shared bin helpers, batch the compact archive append, close test-coverage gaps. PR quality score 9.0.
Adversarial review ran cross-model (Claude + Codex). Claude found a real injection bypass (chat turn-prefixes like
Human:defeating both the write denylist and datamark) plus edge cases; Codex found--allignoringredactand a few mediums. All actionable findings are fixed and tested. Two findings are intentionally left as-is because they fail SAFE (the autopilot pid-reuse and pgrep-substring cases over-refuse sync; tightening them would risk a fail-dangerous path); documented as such.Design Review
No frontend files changed — design review skipped.
Eval Results
Diff-based selection pulls ~142 E2E + 20 LLM-judge tests (the changes touch the shared preamble/resolvers, which fan out to most skills). Paid evals deferred to CI gate-tier; the free suite (gen-skill-docs, parity, goldens, 117 decision/gbrain tests) is green locally.
Plan Completion
Plan: institutional decision-memory (capture → curate → resurface). All three phases delivered — reliable file-only core + bins + Context Recovery; CLAUDE.md section + structured skill emits; optional gbrain
--semantic.find-contradictionsdeferred (the gbrain CLI surface for it can't be verified deterministically yet and the curated-memory source isn't indexed).TODOS
No TODO items completed in this PR.
Test plan
bun testfull suite passes (0 failures)gen:skill-docsclean (0 stale), parity 10/10, ship goldens refreshed🤖 Generated with Claude Code