feat: hardened Google Jules and Cursor automation control plane

[yansigit]

Summary

Adds a hardened, GitHub-native automation control plane coordinating Google Jules, Cursor Cloud Agent, and Cursor Bugbot:

• Google Jules Dispatcher & Reconciler: Trusted workflow () and tested helper () for issue implementation () and plan requests () targeting fork .
• Exact-Head Bugbot Review: Extended PR gating () requiring exact-app and exact-head completed success from Cursor Bugbot, blocking stale/neutral/spoofed checks, with a two-maintainer exact-head waiver ().
• Bounded Repair Loop: Automated repair digest capped at 10 findings / 12 KB, filtering resolved threads and allowing up to 2 repair cycles before escalating to .
• Security Boundaries: Default-branch execution guards on , integration-branch filters on , immutable App ID and user ID verification, and immediate escalation on protected surface expansion.
• Repository Maintenance & Rules: Root rules, fork-targeted , and documentation in .

Verification

• ▶ agent maintenance workflow
  ✔ uses trusted events, reconciliation, and curated schedules (0.527333ms)
  ✔ checks out only trusted default-branch controller code (0.258ms)
  ✔ keeps dispatch bounded, idempotent, and permission gated (0.215792ms)
  ✔ reconciles exact-head reviews and enforces the repair ceiling (0.578125ms)
  ✔ agent maintenance workflow (2.087542ms)
  ▶ maintenance state marker
  ✔ round-trips v1 and fills fields omitted by an early v1 marker (0.953333ms)
  ✔ fails closed for corrupt or invalid state (2.620291ms)
  ✔ cannot terminate its hidden marker through vendor-controlled strings (0.306ms)
  ✔ maintenance state marker (4.437208ms)
  ▶ Cursor Bugbot evidence
  ✔ accepts only a successful exact-name, exact-app, exact-head check (0.393375ms)
  ✔ blocks pending, neutral, stale, and spoofed checks (0.091083ms)
  ✔ does not let an older success mask a newer neutral rerun (0.060042ms)
  ✔ accepts an outage waiver only after two current maintainers approve the exact head (0.177417ms)
  ✔ Cursor Bugbot evidence (0.910208ms)
  ▶ baseline CI evidence
  ✔ requires latest successful exact-head evidence for every configured check (0.128375ms)
  ✔ baseline CI evidence (0.201417ms)
  ▶ controller fail-closed helpers
  ✔ uses the latest labeled or unlabeled event for an active dispatch label (0.110667ms)
  ✔ rejects truncated or malformed GitHub changed-file lists (3.204ms)
  ✔ classifies every documented Jules state without leaving terminal states running (0.13075ms)
  ✔ counts only durable controller state toward the Jules concurrency ceiling (0.092167ms)
  ✔ accepts only Jules-authored fast-forward head movement (0.179209ms)
  ✔ controller fail-closed helpers (3.874083ms)
  ▶ repair findings
  ✔ keeps current-head immutable-author findings and enforces count and byte limits (0.451458ms)
  ✔ repair findings (0.489541ms)
  ▶ quota exhaustion
  ✔ escalates only after the same Jules 429 has lasted 24 hours (0.099333ms)
  ✔ quota exhaustion (0.132541ms)
  ▶ Jules API boundary
  ✔ builds a fork-main automatic-PR request with explicit plan approval (0.114ms)
  ✔ selects only the exact connected GitHub repository source (0.077166ms)
  ✔ retries reads but never blindly retries a create (19.154ms)
  ✔ honors an HTTP-date Retry-After on read-only retries (0.902208ms)
  ✔ fails closed without retrying terminal read or mutation statuses (1.795792ms)
  ✔ follows Jules pagination tokens (0.848959ms)
  ✔ recovers an uncertain create by exact deterministic title without duplicating (1.242167ms)
  ✔ reconciles ambiguous POST 5xx and successful responses with invalid JSON (0.890834ms)
  ✔ rejects an uncertain-create title match from another source (0.311667ms)
  ✔ polls the resource-name tail rather than the opaque session id (0.17425ms)
  ✔ validates that a session output names the live open fork-main PR (0.609459ms)
  ✔ Jules API boundary (26.430458ms)
  ✔ issue automation streams prompts through the digest-pinned Copilot CLI without tool access (2.843708ms)
  ✔ Copilot failures leave issue enforcement and triage retryable (1.06725ms)
  ▶ enforce-pr-target workflow
  ✔ uses pull_request_target without checking out PR head code (1.439083ms)
  ✔ grants contents:write so draft GraphQL mutations work with GITHUB_TOKEN (0.785167ms)
  ✔ fails the required check on a wrong base even if draft conversion fails (0.120541ms)
  ✔ allows main only on the public fork while upstream stays dev-only (0.27675ms)
  ✔ soft-fails ready-for-review restoration the same way (0.206958ms)
  ✔ listens for synchronize so rebase can clear ancestry failures (0.083667ms)
  ✔ uses label events for GUI waivers, hygiene sponsorship, and a trusted CodeRabbit status signal (0.187167ms)
  ✔ wakes from trusted Bugbot checks and revalidates exact-head evidence (0.126459ms)
  ✔ does not add review events that would break the trusted-base model (0.272292ms)
  ✔ queries review threads and feeds them to the findings claim check (0.166125ms)
  ✔ fails closed when review threads cannot be read (0.073834ms)
  ✔ writes exactly one consolidated comment via a single upsert helper (0.134125ms)
  ✔ manages the review-ready status label at the ready moment (0.089083ms)
  ✔ does not embed a literal CodeRabbit review command in the ready notice (0.086834ms)
  ✔ does not rewrite the gate comment when the rebuilt body is unchanged (0.08825ms)
  ✔ keeps CodeRabbit auto-review unfiltered so maintainer PRs are not starved (2.240167ms)
  ✔ migrates legacy two-comment PRs and deletes the old comments (1.356958ms)
  ✔ checks out scripts from the event-specific trusted boundary (never PR head) (1.542875ms)
  ✔ orders same-head CodeRabbit reviews deterministically without timestamps (0.1375ms)
  ✔ loads pr-quality via require from the checked-out scripts (0.060791ms)
  ✔ checks stacked bases via open PR heads before wrong_base enforcement (0.099375ms)
  ✔ strips stale WRONG BRANCH prefix on failure when base is corrected (0.072666ms)
  ✔ enforce-pr-target workflow (11.395333ms)
  ▶ fork auto-release decision
  ✔ dispatches an unused stable package version from green main CI (2.665583ms)
  ✔ skips non-workflow_run events (0.485125ms)
  ✔ skips another triggering workflow (1.25875ms)
  ✔ skips unsuccessful CI (0.173291ms)
  ✔ skips non-main branches (0.116166ms)
  ✔ skips a moved main branch (0.113833ms)
  ✔ skips another package (0.103417ms)
  ✔ skips prerelease versions (0.134417ms)
  ✔ skips versions already on npm (0.130291ms)
  ✔ rejects an empty SHA as malformed input (0.5175ms)
  ✔ rejects a missing package name as malformed input (0.161459ms)
  ✔ fork auto-release decision (6.981417ms)
  ▶ fork auto-release env CLI
  ✔ prints dispatch JSON from env vars without a node heredoc (62.433833ms)
  ✔ treats VERSION_ON_NPM as already published when set (50.352458ms)
  ✔ fork auto-release env CLI (112.944666ms)
  ▶ issue Codex sync did not complete. Fix the reported Codex config issue and retry. lidge-jun/opencodex#1672 regression
  ✔ rejects a reproduction that only echoes the generic final sync failure from Summary (9.325709ms)
  ✔ rejects ordered-list formatting when it only repeats Summary evidence (1.000125ms)
  ✔ normalizes identical multi-line ordered lists on both sides (0.465875ms)
  ✔ preserves numeric failure evidence inside fenced and indented code blocks (1.001417ms)
  ✔ keeps the same failure text valid when Reproduction adds an actionable command (3.313708ms)
  ✔ issue Codex sync did not complete. Fix the reported Codex config issue and retry. lidge-jun/opencodex#1672 regression (15.92675ms)
  ✔ accepts [Bug] Cursor Claude-family models fail via Claude Code path — needs v2.11+ paired trace lidge-jun/opencodex#1162-shaped bug evidence without literal Reproduction/Version/OS headings (10.688583ms)
  ✔ accepts emphasized Environment keys (2.126417ms)
  ✔ does not duplicate case-insensitive reproduction aliases (0.325666ms)
  ✔ does not treat vague alternative headings as actionable reproduction (0.463417ms)
  ✔ still rejects an unknown OpenCodex version from Environment (0.527166ms)
  ✔ still rejects missing OS metadata from Environment (0.195625ms)
  ▶ detectIssueKind
  ✔ detects new feature form without [Feature]: prefix (1.599459ms)
  ✔ detects legacy feature form with [Feature]: prefix (0.09675ms)
  ✔ detects new bug form without [Bug]: prefix (0.111916ms)
  ✔ detects legacy bug form with [Bug]: prefix (0.064459ms)
  ✔ detects provider compatibility form (0.230792ms)
  ✔ detects documentation form (0.087791ms)
  ✔ returns null for unrelated issue with manually applied enhancement label (0.059542ms)
  ✔ uses stored bot kind when headings are removed (0.086583ms)
  ✔ detectIssueKind (5.266667ms)
  ▶ validateIssue - feature
  ✔ keeps nested sub-headings and fenced heading text inside a section ([Bug]: issue-quality closes valid reports because extractSection stops at nested sub-headings lidge-jun/opencodex#541) (8.098ms)
  ✔ ignores markdown headings inside backtick and tilde fences when finding section boundaries ([Bug]: issue-quality closes valid reports because extractSection stops at nested sub-headings lidge-jun/opencodex#541) (0.39775ms)
  ✔ rejects issue [Feature]: Native OpenAI Chat Completions Proxy Endpoint (/v1/chat/completions) lidge-jun/opencodex#208-style duplicate content (0.67825ms)
  ✔ rejects an image-only goal section that hides repeated prose (It is hoped that the usage query will support time-based queries and statistics, as well as key-based queries and statistics lidge-jun/opencodex#1098) (2.538084ms)
  ✔ rejects a markdown-image-only goal section with repeated prose (It is hoped that the usage query will support time-based queries and statistics, as well as key-based queries and statistics lidge-jun/opencodex#1098) (0.505042ms)
  ✔ rejects a markdown image with bracketed alt text in the goal (It is hoped that the usage query will support time-based queries and statistics, as well as key-based queries and statistics lidge-jun/opencodex#1098) (0.362875ms)
  ✔ rejects a markdown image whose URL contains balanced parentheses (It is hoped that the usage query will support time-based queries and statistics, as well as key-based queries and statistics lidge-jun/opencodex#1098) (0.28125ms)
  ✔ preserves a goal section that mixes an image with real text (1.364458ms)
  ✔ treats image/media-only sections as empty via isMediaOnly (1.185083ms)
  ✔ accepts a concise but actionable feature (0.521416ms)
  ✔ rejects issue Change voice chat to different model lidge-jun/opencodex#401-style low-effort feature with placeholder example (0.70825ms)
  ✔ rejects feature reports with placeholder example variants (2.203958ms)
  ✔ reports blank example usage as missing, not placeholder (0.919625ms)
  ✔ accepts a valid legacy feature request without blocker/example headings (0.437459ms)
  ✔ accepts a detailed CJK submission (2.105917ms)
  ✔ rejects terse goal sections that only contain a keyword, digit, or punctuation (1.154833ms)
  ✔ rejects a single long non-CJK word as overly terse (2.4355ms)
  ✔ counts mixed-script CJK text without inflating non-CJK letter length (0.171334ms)
  ✔ rejects mixed-script CJK stubs that only inflate letter counts (0.787042ms)
  ✔ accepts sufficiently detailed goal sections (0.230958ms)
  ✔ accepts a 6-7 word goal when it includes concrete technical detail (0.53375ms)
  ✔ rejects a 6-7 word goal that lacks concrete technical detail (0.514667ms)
  ✔ treats only commands, errors, paths, or exact actions as actionable reproduction detail (4.06375ms)
  ✔ bounds long non-matching reproduction path tokens (9.651ms)
  ✔ rejects fenced placeholder-only examples (1.22725ms)
  ✔ accepts real fenced examples that merely mention N/A (0.393042ms)
  ✔ validateIssue - feature (44.3495ms)
  ▶ validateIssue - bug
  ✔ rejects an empty bug report (0.464416ms)
  ✔ rejects a bug with Summary filled but Reproduction empty (0.19575ms)
  ✔ rejects a bug whose Reproduction is only an ellipsis (Open Codex Error lidge-jun/opencodex#598) (0.136875ms)
  ✔ rejects a bug with Reproduction filled but Summary empty (0.132292ms)
  ✔ accepts a terse real crash report (1.092875ms)
  ✔ accepts a valid legacy bug report without version/OS headings (0.757708ms)
  ✔ accepts a legacy bug with No response in old optional env fields (0.848541ms)
  ✔ accepts a legacy bug with N/A-style placeholders in Version and Operating system (1.6705ms)
  ✔ rejects a new-form bug where env fields were actively cleared (0.581042ms)
  ✔ rejects unknown / don't-know Version values (Unexpected interruption continues to occur lidge-jun/opencodex#624) (2.827167ms)
  ✔ rejects issue Unexpected interruption continues to occur lidge-jun/opencodex#624-style low-effort new-form bug (12.483125ms)
  ✔ rejects a new-form bug with a usable Version but placeholder OS (8.4245ms)
  ✔ rejects a new-form bug when the Version heading was removed (13.806625ms)
  ✔ rejects a new-form bug when the Operating system heading was removed (5.18525ms)
  ✔ rejects a new-form bug whose Reproduction is only a vague phrase (0.516833ms)
  ✔ rejects a Dont work in VSCode Codex plugin lidge-jun/opencodex#977-shaped bug with product keywords but no actionable reproduction (5.63025ms)
  ✔ rejects unknown Operating system stand-ins on the new bug form (2.434208ms)
  ✔ validateIssue - bug (57.82975ms)
  ▶ validateIssue - provider-compatibility
  ✔ rejects when request and response are both absent (0.615084ms)
  ✔ accepts a complete provider compatibility report (1.077583ms)
  ✔ rejects provider compat report when provider/endpoint fields are cleared (2.870375ms)
  ✔ validateIssue - provider-compatibility (5.1205ms)
  ▶ validateIssue - documentation
  ✔ rejects an empty documentation report (0.127167ms)
  ✔ accepts a complete documentation correction (0.207708ms)
  ✔ validateIssue - documentation (0.384458ms)
  ▶ normalisation
  ✔ treats 'No response' as empty (0.244334ms)
  ✔ treats NA and not applicable as placeholders (0.12875ms)
  ✔ detects unusable Version stand-ins without treating them as generic placeholders (0.135666ms)
  ✔ does not treat sentences containing placeholder phrases as empty (0.810125ms)
  ✔ shares one placeholder matcher across clean, isPlaceholder, and isRawPlaceholder (0.142167ms)
  ✔ strips HTML comments (0.064834ms)
  ✔ normalises punctuation and capitalisation (0.1405ms)
  ✔ removes filler phrases (0.154ms)
  ✔ normalisation (2.113916ms)
  ▶ extractSection
  ✔ extracts content between headings (0.129417ms)
  ✔ returns null for missing sections (0.037792ms)
  ✔ extractSection (0.232542ms)
  ▶ shouldReopen
  ✔ allows reopen when timestamps and state match (0.091958ms)
  ✔ forbids reopen when timestamp differs (0.125042ms)
  ✔ forbids reopen when state reason differs (0.070792ms)
  ✔ forbids reopen when bot state is inactive (0.030958ms)
  ✔ returns false when issue is already open (0.024709ms)
  ✔ forbids reopen on maintainer override (0.021292ms)
  ✔ forbids reopen when a human closed the issue (closed_by is not the bot) (0.023291ms)
  ✔ allows reopen when the bot is the recorded closer (0.02625ms)
  ✔ shouldReopen (0.511ms)
  ▶ shouldEnforceClosure
  ✔ enforces when there is no bot state yet (0.038667ms)
  ✔ enforces while the bot still owns an active closure (0.020959ms)
  ✔ does not enforce after a maintainer override (0.020917ms)
  ✔ still enforces after a normal active:false without maintainer override (0.020042ms)
  ✔ shouldEnforceClosure (0.138333ms)
  ▶ translated feature headings and soft-pass
  ✔ accepts Goal / Problem + Expected behaviour as a valid feature (0.317666ms)
  ✔ soft-passes [Feature]: with rich custom headings outside the alias map (0.447583ms)
  ✔ soft-passes retitled feature reports that drop the [Feature]: prefix (0.091958ms)
  ✔ soft-passes retitled bug reports with substantial non-English structure (Claude Desktop 3P Auto Mode classifier retries after 64-token Anthropic OAuth outputs lidge-jun/opencodex#545) (0.309791ms)
  ✔ does not soft-pass a single arbitrary rich heading (Codex fix(ci): soft-pass detailed issues after maintainer retitle lidge-jun/opencodex#564) (0.089334ms)
  ✔ does not soft-pass provider reports that only fill mapped metadata headings (0.170084ms)
  ✔ still rejects empty [Feature]: bodies (0.161625ms)
  ✔ does not treat a title containing problem as a bug (0.037625ms)
  ✔ does not soft-pass long unstructured bodies without headings (0.039042ms)
  ✔ does not classify Expected behaviour + Example as feature without a feature hint (0.061042ms)
  ✔ classifies alias headings as feature when a goal heading is present (0.05675ms)
  ✔ lets a strong bug form override a stale stored feature kind (0.908458ms)
  ✔ accepts US spelling Expected behavior as a behaviour alias (0.370916ms)
  ✔ does not treat enhancement + non-goal aliases as a feature detect hit (0.121708ms)
  ✔ does not let a weak title-prefix detection override stored documentation kind (0.100167ms)
  ✔ translated feature headings and soft-pass (3.895666ms)
  ▶ labelForKind
  ✔ maps kinds to triage labels (0.096792ms)
  ✔ labelForKind (0.128042ms)
  ▶ validateIssue - freeform / non-template
  ✔ rejects a plain freeform body that previously skipped validation (0.265625ms)
  ✔ rejects a 499 client_closed_request errors during web-search tool usage lidge-jun/opencodex#521-shaped Description/Reproduction/Log entry body with a clear message (0.361292ms)
  ✔ still detects and validates a real structured bug as before (0.416625ms)
  ✔ does not treat Summary+Reproduction alone as a bug without prefix or label (0.327542ms)
  ✔ keeps label-backed storedKind validation for enhancement freeform (0.453792ms)
  ✔ validateIssue - freeform / non-template (2.263791ms)
  ▶ rejectsWorkflowDispatchPullRequest
  ✔ rejects pull request numbers on workflow_dispatch (0.111917ms)
  ✔ allows issues and non-dispatch events (0.05125ms)
  ✔ rejectsWorkflowDispatchPullRequest (0.224583ms)
  ▶ rejectsWorkflowDispatchNonDefaultBranch
  ✔ rejects workflow_dispatch runs that are not on the default branch (0.087458ms)
  ✔ allows default-branch dispatches and normal issue events (0.044791ms)
  ✔ rejectsWorkflowDispatchNonDefaultBranch (0.177375ms)
  ▶ mapAreaFieldToLabels
  ✔ maps canonical Area dropdown values (0.89925ms)
  ✔ maps legacy Service lifecycle wording and ignores Other / Multiple areas (0.606291ms)
  ✔ exposes metadata for every non-documentation area label (0.110333ms)
  ✔ mapAreaFieldToLabels (1.69575ms)
  ▶ detectAreaLabels
  ✔ applies Area mapping plus orthogonal heuristics (1.392333ms)
  ✔ adds provider for provider-compatibility form and label (0.957417ms)
  ✔ runs heuristics for Multiple areas / Other without inventing per-provider labels (0.362958ms)
  ✔ does not map Documentation Area onto the documentation kind label (1.292959ms)
  ✔ ignores Operating system metadata for platform heuristics (0.645375ms)
  ✔ uses heuristicBody translation text when Area is Other (1.757959ms)
  ✔ matches truncated streaming wording via truncat stem (0.307ms)
  ✔ detectAreaLabels (7.05275ms)
  ▶ clean() respects fenced code (regression)
  ✔ keeps section text that follows a comment-like literal in a fence (0.085209ms)
  ✔ still strips a real HTML comment outside code (0.03175ms)
  ✔ clean() respects fenced code (regression) (0.145792ms)
  ▶ code-region scanning is GFM-correct and linear (regression)
  ✔ honors a closing fence longer than its opener (0.045708ms)
  ✔ honors a code span containing a line ending (0.026709ms)
  ✔ stays linear on adversarial input (6.597417ms)
  ✔ code-region scanning is GFM-correct and linear (regression) (6.712833ms)
  ▶ hashTranslationSource
  ✔ changes when only the title changes (3.20375ms)
  ✔ changes when only the body changes (0.128875ms)
  ✔ is stable for unchanged title and body (0.082917ms)
  ✔ hashTranslationSource (4.202583ms)
  ▶ splitTranslationBlock
  ✔ handles generated block at end (0.4055ms)
  ✔ preserves suffix after generated block (0.102417ms)
  ✔ preserves prefix before generated block (0.059292ms)
  ✔ does not remove contributor-authored details elsewhere (0.0655ms)
  ✔ fails safely when closing details is missing (0.084875ms)
  ✔ preserves nested details inside translated content via end marker (0.090542ms)
  ✔ removes multi-level nested details only inside the generated block (0.102ms)
  ✔ migrates legacy blocks that close on first details end (0.097625ms)
  ✔ does not greedily erase across duplicate end markers (0.051458ms)
  ✔ splitTranslationBlock (1.258541ms)
  ▶ isPreparedSourceStillCurrent
  ✔ detects body changes between prepare and apply (0.08ms)
  ✔ detects title changes between prepare and apply (0.071458ms)
  ✔ allows apply when only generated translation changed (0.11775ms)
  ✔ isPreparedSourceStillCurrent (0.427041ms)
  ▶ bot-owned control state
  ✔ always includes visible bookkeeping, including English (1.618917ms)
  ✔ incomplete AI/parse failures bookkeep as unknown, never false-English (0.755458ms)
  ✔ English creates a bot comment with visible English bookkeeping (0.440916ms)
  ✔ English updates the canonical bot comment instead of creating duplicates (0.491625ms)
  ✔ updates the oldest sticky control comment even when its state is corrupt (0.673583ms)
  ✔ non-English persist writes or updates a visible bot-owned comment (0.407834ms)
  ✔ ignores author comments containing the control marker (0.216917ms)
  ✔ treats corrupt control state as missing (0.16125ms)
  ✔ never treats the issue body as authoritative control state (0.403375ms)
  ✔ failed sticky update preserves existing control comments and does not delete (1.260458ms)
  ✔ re-fetches comments when none are supplied and still verifies authorship (0.345041ms)
  ✔ failed comment update preserves the previous comment (0.290542ms)
  ✔ deletes redundant bot comments only after sticky replacement succeeds (0.159708ms)
  ✔ cleanup failure leaves valid fallback comments intact (0.175292ms)
  ✔ cooldown and hourly limits survive repeated issue events via bot comments (0.20075ms)
  ✔ rapid sequential edits cannot invoke the model repeatedly (0.107125ms)
  ✔ persistence never mutates the issue title or body (0.066334ms)
  ✔ selects only github-actions control comments (0.087833ms)
  ✔ reader and selector agree on the newest control comment (0.064334ms)
  ✔ round-trips base64url control state without HTML breakout (0.610709ms)
  ✔ rejects invalid decoded payloads (0.233ms)
  ✔ scrubs injection characters from detected language (0.14575ms)
  ✔ heals slightly-future prior within skew and uses wall-clock attemptedAt (0.208417ms)
  ✔ rejects far-future attemptedAt and strips far-future recent entries (0.147125ms)
  ✔ corrupt far-future comment does not outrank a valid current comment (0.724917ms)
  ✔ mergeTranslationAttemptState ignores poisoned far-future prior (0.065833ms)
  ✔ canonicalisation preserves valid bounded recent history across comments (0.083333ms)
  ✔ strips trailing obsolete markers without treating them as control state (0.046125ms)
  ✔ preserves author whitespace when stripping a trailing obsolete marker (0.057292ms)
  ✔ leaves marker-like author content inside fences, quotes, and prose untouched (0.033208ms)
  ✔ translation application preserves marker-like author content (0.0435ms)
  ✔ stale-source checking detects actual author edits after safe normalisation (0.487292ms)
  ✔ rate limits repeated non-ASCII detections (0.120333ms)
  ✔ defuses mention-shaped tokens without rewriting emails or mid-token at-signs (0.061625ms)
  ✔ ignores forged body-embedded legacy state (0.088125ms)
  ✔ deleteVerifiedControlComments skips author-forged marker comments (0.121208ms)
  ✔ bot-owned control state (11.903792ms)
  ▶ eligibility
  ✔ allows meaningful title with short body (0.0495ms)
  ✔ re-translates after title-only edits (0.044083ms)
  ✔ eligibility (0.133458ms)
  ▶ appendTranslationBlock
  ✔ replaces an existing generated block exactly once (0.05575ms)
  ✔ truncates translations that would exceed the issue body limit (0.153ms)
  ✔ appendTranslationBlock (0.260875ms)
  ▶ issue comment translation
  ✔ skips bots, control comments, and pull-request threads (0.15775ms)
  ✔ hashes comments under comment: so they do not collide with issue hashes (0.048916ms)
  ✔ allows a meaningful non-English comment and builds an in-place translation block (0.205166ms)
  ✔ rejects comments without a usable numeric id (0.057875ms)
  ✔ skips unchanged comment bodies via source hash (0.079458ms)
  ✔ keeps issue and comment completed hashes independent (0.084959ms)
  ✔ enforces minSourceChars on stripped comment body only (0.034666ms)
  ✔ short comments do not become eligible via comment: namespace padding (0.033792ms)
  ✔ stale-guards comment edits with isPreparedSourceStillCurrent (0.035125ms)
  ✔ issue comment translation (0.819417ms)
  ▶ missingRequiredTranslationFields
  ✔ requires translated title/body only when the matching source field is nonempty (0.067083ms)
  ✔ missingRequiredTranslationFields (0.0865ms)
  ▶ sourceComplete vs rate-limit attempts
  ✔ incomplete attempts count toward rate limits but stay retryable after cooldown (0.100542ms)
  ✔ completed no-translation and successful apply mark the source hash (0.064833ms)
  ✔ incomplete issue and comment attempts preserve a prior completed hash (0.106708ms)
  ✔ sourceComplete vs rate-limit attempts (0.314208ms)
  ▶ deterministic duplicate auto-close
  ✔ does not treat the generic Codex sync did not complete. Fix the reported Codex config issue and retry. lidge-jun/opencodex#1672 final sync message as duplicate proof (4.312708ms)
  ✔ selects an AI-nominated duplicate when both reports share an exact specific failure signature (0.803125ms)
  ✔ does not treat a semantic version as a structured field path (0.138334ms)
  ✔ still accepts an exact IPv4 address as a discriminator (0.103375ms)
  ✔ keeps case-distinct file paths distinct (0.159292ms)
  ✔ still matches a genuine structured field path (0.091875ms)
  ✔ recognizes fails as a strong failure signal (0.08125ms)
  ✔ recognizes short-prefix named error classes as specific duplicate evidence (0.0735ms)
  ✔ selects the longest shared signature across all nominated candidates (0.10625ms)
  ✔ preserves technical punctuation so distinct signatures cannot collapse together (0.097333ms)
  ✔ never selects the current issue as its own duplicate (0.070833ms)
  ✔ uses a locale-independent code-unit tie-breaker for equal-length signatures (0.077958ms)
  ✔ does not auto-close an AI duplicate without an exact strong failure signature (0.063209ms)
  ✔ ignores exact strong matches that the AI did not nominate as duplicates (0.043916ms)
  ✔ does not promote short generic HTTP failures to auto-close signatures (0.045417ms)
  ✔ workflow searches open and recently closed issues and closes only through duplicate state reason (0.356958ms)
  ✔ deterministic duplicate auto-close (8.540125ms)
  ▶ hasConcreteRelatedSignature
  ✔ rejects standalone ECONNRESET referring only to the new issue (0.951458ms)
  ✔ rejects shared route with different failures (0.651667ms)
  ✔ rejects same provider with different root causes (0.262208ms)
  ✔ does not treat ordinary e-words as error constants (0.573625ms)
  ✔ does not treat capitalized common words as errno tokens (0.381708ms)
  ✔ ignores bare three-digit numbers without status context (0.105708ms)
  ✔ rejects different HTTP statuses despite component overlap (0.079333ms)
  ✔ rejects distinct statuses even when the shared-comparison gate is open (0.058375ms)
  ✔ rejects shared wording that borrows a concrete token from only one issue (0.980958ms)
  ✔ rejects mixed HTTP and errno failures attributed to different issues (0.265083ms)
  ✔ rejects this/current/present issue attributions after generic both-wording (0.827416ms)
  ✔ keeps shared concrete failure signatures (0.345292ms)
  ✔ keeps shared errno plus shared comparison (0.124375ms)
  ✔ hasConcreteRelatedSignature (8.199167ms)
  ▶ hardenRelatedMatches
  ✔ drops 503 from /v1/responses via Codex even though direct proxy curl to same endpoint returns 200 OK (Xiaomi MiMo provider, openai-chat adapter) lidge-jun/opencodex#452-style weak related links to unrelated HTTP errors (0.607125ms)
  ✔ drops generic same-client / same-app overlap without a concrete signature (0.066042ms)
  ✔ keeps related when shared comparison has a concrete signature (0.118375ms)
  ✔ one valid related entry does not validate other weak entries (0.066042ms)
  ✔ each related entry requires its own concrete reason (0.050666ms)
  ✔ caps related at 3 and never overlaps duplicates (0.161291ms)
  ✔ hardenRelatedMatches (3.292875ms)
  ▶ parseTriageMatches
  ✔ returns null when hardening clears a weak related-only match (0.20825ms)
  ✔ keeps strong duplicates even if related is weak (0.083417ms)
  ✔ ignores unknown and self issue numbers (0.103292ms)
  ✔ ignores malformed related objects and legacy string related arrays (0.114291ms)
  ✔ prevents duplicate and related lists from overlapping (0.060541ms)
  ✔ caps related output at 3 (0.069792ms)
  ✔ parseTriageMatches (0.702959ms)
  ▶ parseAiJson
  ✔ strips a json fence before parsing (0.096416ms)
  ✔ returns null for unparseable input (0.088875ms)
  ✔ parseAiJson (0.2365ms)
  ▶ sanitizeReason
  ✔ strips markdown and mention markers (0.0725ms)
  ✔ sanitizeReason (0.123375ms)
  ▶ parseAiResponse
  ✔ parses plain JSON objects (0.752042ms)
  ✔ parses fenced JSON (0.137833ms)
  ✔ rejects invalid JSON, empty input, and arrays (0.102375ms)
  ✔ repairs JS-style apostrophe escapes that models put in JSON strings (0.078625ms)
  ✔ repairs apostrophe escapes inside a fenced code block (0.251334ms)
  ✔ preserves literal backslashes for non-apostrophe invalid escapes (0.118833ms)
  ✔ repairs invalid or truncated unicode escapes instead of keeping bare \u (1.1195ms)
  ✔ keeps requires_translation=true for issue and comment payloads with invalid escapes (88.231541ms)
  ✔ parseAiResponse (91.444084ms)
  ▶ parse-issue-translation-response process
  ✔ writes translated outputs for requires_translation=true (28.749958ms)
  ✔ writes English path without translation fields (28.795208ms)
  ✔ marks invalid JSON incomplete so sources stay retryable (87.426ms)
  ✔ leaves translate decisions incomplete until apply succeeds (27.34475ms)
  ✔ treats only boolean false as a completed no-translation decision (35.516ms)
  ✔ keeps invalid requires_translation values incomplete and retryable (244.259875ms)
  ✔ treats string requires_translation as incomplete, not English-complete (29.945208ms)
  ✔ supports multiline bodies and shell-looking content (28.778ms)
  ✔ writes false outputs on invalid JSON and empty responses (95.924958ms)
  ✔ parses fenced responses in the process path (31.716375ms)
  ✔ scrubs control characters from language and title (30.536334ms)
  ✔ parse-issue-translation-response process (669.391167ms)
  ▶ patch parsing
  ✔ returns added content without diff headers (1.466084ms)
  ✔ detects empty catch blocks across added lines (0.476208ms)
  ✔ keeps hunk context and added lines for result scanning (0.251208ms)
  ✔ patch parsing (3.447167ms)
  ▶ assessHygiene
  ✔ requires regression coverage for behavior changes (0.947334ms)
  ✔ accepts behavior changes with tests or approved exception (0.206042ms)
  ✔ does not read an empty catch across a hunk boundary (0.085ms)
  ✔ treats a deleted lockfile as no dependency change (0.098792ms)
  ✔ does not demand a test for a comment-only source change (0.177792ms)
  ✔ still demands a test when a comment change carries any code (0.6255ms)
  ✔ classifies renamed behavior files on both sides (0.471917ms)
  ✔ accepts a renamed behavior file when tests are included (1.157792ms)
  ✔ classifies renamed generated files on both sides (0.16575ms)
  ✔ blocks added suppressions (0.219125ms)
  ✔ blocks focused or skipped tests (0.183917ms)
  ✔ blocks empty catches (0.128667ms)
  ✔ detects a catch emptied by deletion (0.107125ms)
  ✔ does not flag a nonempty catch in a hunk with unrelated deletions (0.102209ms)
  ✔ blocks generated output and orphan lockfile churn (0.109209ms)
  ✔ allows removal of generated output (0.079625ms)
  ✔ does not count deleted tests as regression coverage (0.073542ms)
  ✔ allows maintainer-approved narrow exceptions (0.085041ms)
  ✔ assessHygiene (5.794583ms)
  ▶ collectDeterministicHygieneFailures
  ✔ combines patch hygiene and sponsored-surface failures (0.299291ms)
  ✔ skips sponsorship for maintainers with push permission (0.099792ms)
  ✔ requires sponsorship when renaming away from a restricted path (0.195625ms)
  ✔ exposes hints and gate labels for the Ready coupling (0.089292ms)
  ✔ collectDeterministicHygieneFailures (0.851875ms)
  ▶ pr-hygiene workflow trust boundary
  ✔ checks out trusted scripts from an integration branch, never a PR-controlled ref (0.538083ms)
  ✔ never lets a PR-controlled ref reach an executable step (0.831916ms)
  ✔ uses repository permission level for the sponsorship exemption (0.193959ms)
  ✔ pr-hygiene workflow trust boundary (1.681292ms)
  ▶ detectTypeLabelFromTitle
  ✔ maps conventional prefixes to type labels (1.004375ms)
  ✔ maps sentence-case prefixes when no conventional colon is present (0.329292ms)
  ✔ returns null without a recognized prefix (0.16125ms)
  ✔ detectTypeLabelFromTitle (2.30075ms)
  ▶ hasHumanTypeLabelOverride
  ✔ is false when only the Actions bot touched type labels (0.263417ms)
  ✔ is true after a human replaces the bot type label (PR fix(codex): warn or restart stale app-server after sync lidge-jun/opencodex#518) (0.198542ms)
  ✔ stays true even if the bot later reverts the human choice (0.149917ms)
  ✔ ignores non-type labels from humans (0.1355ms)
  ✔ hasHumanTypeLabelOverride (1.118875ms)
  ▶ planTypeLabelSync
  ✔ adds the detected label and removes other type labels when bot-owned (1.036666ms)
  ✔ is a no-op add when the detected label is already present (0.099917ms)
  ✔ skips when a human has overridden the type label (0.0945ms)
  ✔ labels sentence-case bug-fix titles (PR Fix Console Go tool schema sanitization lidge-jun/opencodex#524) (0.064041ms)
  ✔ skips titles without a recognized prefix (0.066291ms)
  ✔ planTypeLabelSync (1.548292ms)
  ▶ detectTypeLabelFromCommits
  ✔ reads the type from unanimous commits (0.582417ms)
  ✔ treats chore as supporting, not competing (PR stack 5/7: probe reset-derived cooldowns without waiting to be selected (#915) lidge-jun/opencodex#955 shape) (0.133459ms)
  ✔ keeps chore when nothing else competes (0.125167ms)
  ✔ abstains on a genuine mix of fix and feat (0.069791ms)
  ✔ reads only the first line of a multi-line commit message (0.068583ms)
  ✔ returns null for absent or unusable input (0.061333ms)
  ✔ detectTypeLabelFromCommits (1.193167ms)
  ▶ planTypeLabelSync commit fallback
  ✔ labels a stack PR whose title carries no type (0.149042ms)
  ✔ does not let commits override a title that already classifies (0.075ms)
  ✔ still skips when neither the title nor the commits classify (0.073667ms)
  ✔ still honours a human override before consulting commits (0.078125ms)
  ✔ planTypeLabelSync commit fallback (0.503583ms)
  ▶ pr-labeler workflow
  ✔ listens for labeled and unlabeled so human overrides cancel stale sync runs (0.289416ms)
  ✔ keeps trusted default-branch checkout, concurrency cancel, and minimal permissions (0.448125ms)
  ✔ pr-labeler workflow (0.82375ms)
  ▶ parseMaintainerLogins
  ✔ reads the current-maintainers table and excludes the change log (1.51ms)
  ✔ returns an empty list when the section heading is missing (0.296542ms)
  ✔ does not match a ### subsection or prose mentioning the heading (0.152584ms)
  ✔ accepts a valid CRLF heading with trailing whitespace and a following H2 (0.137375ms)
  ✔ handles empty and duplicate-free output (0.149708ms)
  ✔ parseMaintainerLogins (3.487458ms)
  ▶ inlineCode
  ✔ wraps values with a delimiter longer than any backtick run (0.609084ms)
  ✔ inlineCode (2.385625ms)
  ▶ readinessChecklistLines
  ✔ mirrors per-item checked state (0.425375ms)
  ✔ readinessChecklistLines (0.663125ms)
  ▶ buildGateCommentBody
  ✔ carries the marker, serialized state, status, mirror, and tick count (0.420375ms)
  ✔ renders a ready status without a checklist when not required (0.167292ms)
  ✔ does not claim ready when the PR is kept in draft (0.156625ms)
  ✔ buildGateCommentBody (0.972375ms)
  ▶ descriptionFailureLines
  ✔ covers every reason (0.214958ms)
  ✔ descriptionFailureLines (0.354083ms)
  ▶ buildFailureSections
  ✔ builds a wrong-base section naming every allowed base (0.329041ms)
  ✔ builds ancestry, description, and screenshot sections (0.190625ms)
  ✔ buildFailureSections (0.668667ms)
  ▶ failureSummary
  ✔ names every failure kind (0.129583ms)
  ✔ failureSummary (0.198709ms)
  ▶ buildStaleNotice
  ✔ names the recorded and current heads when a completion drifted (0.159125ms)
  ✔ covers the never-recorded predate case (0.042334ms)
  ✔ covers an unrecorded complete checklist on synchronize (0.038458ms)
  ✔ matches the injected section text it resets (0.05975ms)
  ✔ buildStaleNotice (0.35775ms)
  ▶ buildClaimCheckNotice
  ✔ names the latest-dev violation and the reset action (0.066375ms)
  ✔ ignores a stale ci_green code without inventing GitHub-CI copy (0.040917ms)
  ✔ returns only the reset line for an empty violation list (0.362917ms)
  ✔ buildClaimCheckNotice (0.510916ms)
  ▶ buildFindingsClaimNotice
  ✔ names each bot with unresolved threads and the untick (0.074875ms)
  ✔ handles a single bot with one thread (0.030458ms)
  ✔ buildFindingsClaimNotice (0.135625ms)
  ▶ hygiene section round-trip
  ✔ renders a hygiene block in the gate comment when requested (0.075041ms)
  ✔ extracts the hygiene content from a gate comment (0.1465ms)
  ✔ replaces an existing hygiene block without duplicating it (0.056583ms)
  ✔ appends a hygiene block when the gate comment has none (0.056709ms)
  ✔ ignores delimiter text embedded inside a hygiene content line (0.216666ms)
  ✔ preserves both sections across an interleaved gate rebuild and hygiene update (0.120833ms)
  ✔ hygiene section round-trip (0.782875ms)
  ▶ durable CodeRabbit outside-diff findings
  ✔ uses CodeRabbit's stable cr-comment markers as finding identities (1.196625ms)
  ✔ keeps standalone outside-diff findings active without an inline thread (0.238417ms)
  ✔ adds outside-diff markers without double-counting the review actionable total (0.164375ms)
  ✔ a later clean CodeRabbit review on the same head clears older markers (1.062917ms)
  ✔ uses review id as a deterministic tie-breaker when timestamps are missing (0.128875ms)
  ✔ ignores CodeRabbit markers from an older head and from human reviews (0.068292ms)
  ✔ deduplicates repeated markers in the review body (0.111083ms)
  ✔ durable CodeRabbit outside-diff findings (3.960708ms)
  ▶ enforcer state markers
  ✔ parses a valid enforcer state marker (1.9785ms)
  ✔ returns null for markerless or unreadable state and warns (0.584125ms)
  ✔ round-trips through stateMarker (0.130333ms)
  ✔ parses and serializes readiness state with warnings on failure (0.203292ms)
  ✔ enforcer state markers (3.980583ms)
  ▶ state defaults
  ✔ builds the cleared enforcer state (0.096792ms)
  ✔ builds the fresh active enforcer state (0.053417ms)
  ✔ builds the fresh readiness state at the current version (0.05725ms)
  ✔ state defaults (0.311208ms)
  ▶ completionIsStale
  ✔ is not stale when the recorded completion head matches the live head (0.911209ms)
  ✔ is stale when the recorded head differs from the live head, even with an open checklist (0.17125ms)
  ✔ is stale when ticks predate the live head on a first completion (0.16275ms)
  ✔ is not stale when ticks predate the live head but nothing is ticked (0.111875ms)
  ✔ is stale when a complete checklist has no recorded head on synchronize (0.086958ms)
  ✔ is not stale for an unrecorded complete checklist on a non-synchronize event (0.0685ms)
  ✔ is stale when the event delivered no head SHA at all (issue_comment rerun) (0.076ms)
  ✔ is not stale for maintainers or absent checklists (0.073916ms)
  ✔ completionIsStale (1.962916ms)
  ▶ readinessClaimViolations
  ✔ passes when the head is current enough (0.221084ms)
  ✔ never treats local CI as a bot-verifiable claim (0.109625ms)
  ✔ flags a head more than the threshold behind the base (0.081542ms)
  ✔ fails closed when the behind count is unknown (0.063666ms)
  ✔ honours a custom threshold (0.058333ms)
  ✔ readinessClaimViolations (0.693125ms)
  ▶ unresolvedFindingsClaim
  ✔ passes when there are no review threads at all (0.421541ms)
  ✔ passes when every bot thread is resolved (0.208208ms)
  ✔ flags one unresolved Codex thread and counts it per bot (0.207291ms)
  ✔ flags unresolved threads from both bots and counts each (0.173417ms)
  ✔ ignores unresolved threads from humans (0.080917ms)
  ✔ fails closed on a thread with no resolution state (0.0725ms)
  ✔ exposes the bot allowlist (0.055333ms)
  ✔ unresolvedFindingsClaim (1.409875ms)
  ▶ coderabbitOutsideDiffFindings
  ✔ flags a CodeRabbit review of the live head with actionable comments (0.1855ms)
  ✔ ignores a review of a different head (0.081042ms)
  ✔ ignores a review reporting zero actionable comments (0.073916ms)
  ✔ uses the most recent review of the live head (0.13825ms)
  ✔ returns clean for no reviews or no live head (0.116666ms)
  ✔ ignores a human review that quotes the actionable-comments line (0.076583ms)
  ✔ sorts undated reviews last deterministically (0.082125ms)
  ✔ coderabbitOutsideDiffFindings (0.908083ms)
  ▶ unresolvedFindingsClaim with outside-diff supplement
  ✔ does not count outside-diff when no unresolved bot thread exists (0.196584ms)
  ✔ adds the outside-diff count to an unresolved thread count (0.181583ms)
  ✔ keeps a resolved thread set clean even with a stale review (0.117167ms)
  ✔ unresolvedFindingsClaim with outside-diff supplement (0.645917ms)
  ▶ gate state
  ✔ round-trips through gateStateMarker and parseGateState (0.273959ms)
  ✔ returns null for markerless or unreadable gate state and warns (0.19325ms)
  ✔ builds a fresh gate state (0.103417ms)
  ✔ merges legacy enforcer + readiness states (0.15025ms)
  ✔ keeps enforcer-owned auto-draft when the readiness record says false (0.077875ms)
  ✔ migrates with either legacy state absent (0.070791ms)
  ✔ gate state (1.0665ms)
  ▶ isWrongAncestry
  ✔ flags fix(windows): follow active Codex home for tray listener lidge-jun/opencodex#644-shaped compares (0 behind main, far behind base, few ahead of main) (0.315542ms)
  ✔ uses threshold 20 by default (0.058875ms)
  ✔ passes when head is behind main (not sitting on main tip) (0.043541ms)
  ✔ passes stale dev-based branches that are many commits ahead of main (0.163583ms)
  ✔ isWrongAncestry (1.49125ms)
  ▶ authorHasPushPermission
  ✔ accepts write/maintain/admin only (0.245917ms)
  ✔ authorHasPushPermission (0.297541ms)
  ▶ assessPrDescription
  ✔ rejects empty and comment-only bodies (1.766958ms)
  ✔ rejects placeholder-only bodies (0.270792ms)
  ✔ rejects literal escaped newlines like fix(windows): follow active Codex home for tray listener lidge-jun/opencodex#644 (0.056459ms)
  ✔ rejects thin real-newline bodies (0.108291ms)
  ✔ rejects an untouched GitHub PR template as empty/thin (0.143875ms)
  ✔ accepts two rich markdown sections (0.235375ms)
  ✔ accepts unstructured bodies that are long enough with multiple blocks (0.142667ms)
  ✔ assessPrDescription (2.889209ms)
  ▶ hasGuiCue
  ✔ matches gui as a whole word, case-insensitively, in title or body (0.270291ms)
  ✔ does not match negated gui phrases (0.0415ms)
  ✔ does not match gui inside other words (0.034375ms)
  ✔ does not match missing or non-string inputs (0.027917ms)
  ✔ hasGuiCue (0.431625ms)
  ▶ guiPathsChanged
  ✔ matches gui/ paths with a slash guard (0.075875ms)
  ✔ guiPathsChanged (0.131959ms)
  ▶ isChangedFileListTruncated
  ✔ treats head drift, invalid counts, and oversized lists as truncated (0.125417ms)
  ✔ isChangedFileListTruncated (0.170833ms)
  ▶ hasGuiOverride
  ✔ matches a maintainer comment with a negation phrase (0.093209ms)
  ✔ does not let the PR author or a non-collaborator waive the gate (0.182541ms)
  ✔ does not match a comment that names gui without negating it (0.098042ms)
  ✔ does not match a negation that belongs to another sentence or line (0.082833ms)
  ✔ is clean for no comments or a comment without a body (0.071917ms)
  ✔ hasGuiOverride (0.704958ms)
  ▶ hasScreenshotEvidence
  ✔ accepts embedded markdown images (0.394166ms)
  ✔ accepts HTML img tags (0.09425ms)
  ✔ accepts reference-style images with a matching definition (0.178541ms)
  ✔ rejects image syntax inside fenced code and HTML comments (0.041625ms)
  ✔ rejects img tags without a renderable src and references without a definition (0.038ms)
  ✔ rejects plain links, bare image URLs, and text-only bodies (0.028167ms)
  ✔ hasScreenshotEvidence (0.872625ms)
  ▶ review readiness checklist
  ✔ builds exactly the four required boxes inside the markers (0.093167ms)
  ✔ keeps the closing 'ready for review' box separated by a blank line (0.041042ms)
  ✔ reports absent when the body has no markers (0.416458ms)
  ✔ counts checked boxes and requires all four for completion (0.151166ms)
  ✔ reports per-item state so the mirror marks the right boxes (0.046958ms)
  ✔ treats a reworded but complete section as complete (0.038625ms)
  ✔ stays incomplete for fewer or extra boxes inside the markers (0.047833ms)
  ✔ treats inverted or partial markers as present-but-incomplete, never appends again (0.075583ms)
  ✔ appends once and is idempotent (0.058167ms)
  ✔ appends cleanly to an empty body (0.03925ms)
  ✔ strips the marker-bounded section and leaves the rest intact (0.05225ms)
  ✔ strips a section-only body to empty and leaves markerless bodies alone (0.028084ms)
  ✔ resets every checked box to unticked and keeps the surrounding body (0.065083ms)
  ✔ resets a partially ticked section as well (0.035208ms)
  ✔ preserves the surrounding author formatting exactly (0.034083ms)
  ✔ is idempotent on an already-unticked section (0.033ms)
  ✔ leaves markerless and malformed bodies alone (0.027459ms)
  ✔ review readiness checklist (1.432541ms)
  ▶ uncheckReviewReadinessBoxes
  ✔ unchecks only the requested boxes (0.117583ms)
  ✔ can uncheck several boxes at once (0.037542ms)
  ✔ preserves the surrounding author content exactly (0.037583ms)
  ✔ is idempotent on an already-unchecked box (0.042125ms)
  ✔ leaves markerless and malformed bodies alone (0.428875ms)
  ✔ uncheckReviewReadinessBoxes (0.730625ms)
  ▶ assessPrDescription with the readiness section
  ✔ never counts the injected checklist as description substance (0.1685ms)
  ✔ assessPrDescription with the readiness section (0.195583ms)
  ▶ collectPrQualityFailures
  ✔ reports wrong_base without requiring ancestry inputs (0.187375ms)
  ✔ reports wrong_base and bad_description together for main + empty body (0.041541ms)
  ✔ reports wrong_ancestry for contributor on fix(windows): follow active Codex home for tray listener lidge-jun/opencodex#644-shaped compare (0.079542ms)
  ✔ skips ancestry for push permission but still flags bad description (0.034791ms)
  ✔ applies ancestry when permission lookup failed (fail closed) (0.061333ms)
  ✔ does not flag stale dev-based branches that are far ahead of main (0.05425ms)
  ✔ skips ancestry when compare lookup failed (cannot evaluate) (0.053417ms)
  ✔ skips wrong_base when stackedBase is set (0.050042ms)
  ✔ still flags wrong_base for non-allow-list bases without stackedBase (0.039958ms)
  ✔ flags gui/ file changes without a screenshot (0.069166ms)
  ✔ does not flag a gui title when no gui/ files changed (0.051ms)
  ✔ flags truncated file lists even when gui/ is not in the partial list (0.353958ms)
  ✔ flags truncated file lists when gui/ appears in the partial list (0.322875ms)
  ✔ does not flag no gui changes text without gui/ file changes (0.073208ms)
  ✔ flags a gui mention in the body without a screenshot when gui/ changed (0.305208ms)
  ✔ waives the screenshot gate for a maintainer override comment (0.064209ms)
  ✔ keeps the screenshot gate when only the PR author claims no gui change (0.046792ms)
  ✔ accepts a gui title when a screenshot image is embedded (0.108166ms)
  ✔ accepts a gui title when the screenshot uses reference-style markdown (0.173833ms)
  ✔ still flags gui/ changes when image syntax is only inside a code fence (0.155292ms)
  ✔ ignores the template's own gui/screenshot instruction (0.062834ms)
  ✔ collectPrQualityFailures (2.572458ms)
  ▶ comment stripping respects fenced code (regression)
  ✔ keeps a screenshot that follows a comment-like literal in a fence (0.034541ms)
  ✔ still ignores a screenshot inside a real HTML comment (0.022625ms)
  ✔ comment stripping respects fenced code (regression) (0.083ms)
  ▶ isRestrictedPath
  ✔ covers auth, workflow, release, and dependency surfaces (0.366625ms)
  ✔ leaves ordinary product surfaces alone (0.264083ms)
  ✔ isRestrictedPath (1.382625ms)
  ▶ isAgentProtectedPath
  ✔ escalates optional-Lab boundaries, provider code, and rename sources (0.135625ms)
  ✔ allows the two scheduled maintenance surfaces (0.90875ms)
  ✔ isAgentProtectedPath (1.183209ms)
  ▶ assessSponsoredSurface
  ✔ requires sponsorship for a restricted surface (1.240583ms)
  ✔ passes once a maintainer sponsors it (0.218875ms)
  ✔ exempts an author who can already push (0.154708ms)
  ✔ applies to every contributor, not only first-timers (0.138583ms)
  ✔ ignores a pull request that touches nothing restricted (0.179708ms)
  ✔ assessSponsoredSurface (2.223833ms)
  ▶ release dispatch guard
  ✔ accepts an exact audited SHA on main (0.400542ms)
  ✔ accepts an exact audited SHA on preview (0.114292ms)
  ✔ rejects non-workflow_dispatch events (0.094292ms)
  ✔ rejects release dispatches from unapproved refs (0.052167ms)
  ✔ requires expected-sha (0.050292ms)
  ✔ requires a full 40-character commit SHA (0.049ms)
  ✔ rejects when the selected ref moved after audit (0.061333ms)
  ✔ release dispatch guard (1.803125ms)
  ✔ streams a large prompt over stdin and maps the Copilot token to GITHUB_TOKEN (300.328875ms)
  ✔ surfaces Copilot stderr and preserves a non-zero exit code (216.517708ms)
  ✔ kills a hung Copilot process at the configured timeout (113.006916ms)
  ℹ tests 552
  ℹ suites 94
  ℹ pass 552
  ℹ fail 0
  ℹ cancelled 0
  ℹ skipped 0
  ℹ todo 0
  ℹ duration_ms 913.518709 (552 passing tests)
• bun test v1.3.14 (0d9b296a) (227 passing tests)
• bun install v1.3.14 (0d9b296a)

Checked 6 installs across 26 packages (no changes) [12.00ms]

• Privacy scan passed
• Workflow YAML parsing and script syntax compilation checks
• Dedicated security review and audit performed with subagents (GPT-5.6 Luna)

Checklist

• Scope stays focused and avoids unrelated cleanup.
• Docs or release notes were updated when needed.
• Security-sensitive changes were reviewed for secrets, auth, and unsafe defaults.

[github-actions]

✅ Deterministic PR hygiene checks passed.