release: baton 1.3.2 — interactive-path enforcement engages; single-source run-trail count#35
Merged
Merged
Conversation
…dger/seam fixes (candidate) Live interactive /baton findings, fixed forward. HELD — no version bump/push. ROOT-CAUSE FIX (biggest): disposition_gate read only the PROJECT .claude/settings.json to decide whether the spawn/triage sidecars were wired — but the INTERACTIVE path wires them in the USER-GLOBAL ~/.claude/settings.json. In an ordinary repo (no project settings), the deriver therefore believed the sidecars were unwired and silently fell back to record-only: the completeness gate went SILENT and the anti-forgery spawn-match no-op'd (observed live as _sidecar_diag.enforcement_active:false with a real security-review spawn in lane_spawns.jsonl). Now sidecar_enforcement_active / triage_sidecar_active consult project + project-local + user-global settings, so both gates engage on the interactive path. Tests pass explicit paths (checked alone); a new selftest drives the global-only scenario; hooks-e2e runs with a hermetic HOME so "unwired" cases don't read the dev's real global config. - The reconciliation binds a claimed specialist by subagent_type NAME (token-equality), so a genuine lane matches and a claimed specialist with no real spawn of that type fails to UNVERIFIED-SEAM; task_id null is not the blocker. SKILL.md documents the contract_lane format (cite subagent_type; binds by type; per-instance binding is the post-hoc scorer's job). 1.3.2 ledger/seam fixes: - ledger count: OWN lines only (supersedes 1.3.1 max), anchored to the lane-line shape — count can't contradict visible lines in either direction. - record_seam.py (new): manager-run recorder for an inline-triaged seam -> arms the completeness gate (add-only). SKILL.md/triage.md: inline triage is for non-seam work; a named seam MUST be recorded. - NO close-out WARN: a reliable transcript-scan is impossible (skill's own seam examples + TRIAGE-SEAMS: none are in the transcript) -> removed per review; post-hoc scorer is the authority. - close-out relabel "this session" -> "so far"; task_id documented expected-null; doctor decouple comment hardened; record_seam_test gated in CI; openspec add-infra-seam-taxonomy (validates --strict). Verified: full smoke green (8 hook self-tests, hooks-e2e 22); fresh N-lane acceptance; interactive-path enforcement selftest; independent cold-read review (3 findings resolved). Co-Authored-By: Claude Opus 4.8 <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
1 participant
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.
Patch release (1.3.1 → 1.3.2). Version bump + CHANGELOG
## 1.3.2+ the fixes below. Pushing thev1.3.2tag after merge triggersrelease.yml.The significant fix — interactive-path enforcement now engages
The close-out deriver decided whether the spawn/triage sidecars were wired by reading only the project
.claude/settings.json. The interactive/batonpath wires the enforcement hooks in the user-global~/.claude/settings.json, so in an ordinary repo (no project settings) the deriver believed the sidecars were unwired and silently fell back to record-only — the completeness gate went silent and the anti-forgery spawn-match no-op'd, even with a real independent lane recorded (observed live:_sidecar_diag.enforcement_active: falsewith a realsecurity-reviewspawn inlane_spawns.jsonl). The wired-check now consults project + project-local + user-global settings, so both gates engage on the interactive path. Verified live:sidecar_enforcement_active()/triage_sidecar_active()returnTruefrom a bare cwd where they wereFalse.Other fixes
hooks/record_seam.py(manager-run) writes an inline-named seam to the sametriaged_seams.jsonlthe completeness gate reads (add-only). SKILL.md/triage.md: inline triage is for non-seam work; a named seam must be machine-recorded.max()reconciliation), anchored to the lane-line shape — the count can't contradict the visible lines in either direction. Close-out relabel "this session" → "so far".task_iddocumented expected-null (specialist match binds bysubagent_type).contract_laneformat documented (citesubagent_type; binds by type; per-instance is the post-hoc scorer's job).TRIAGE-SEAMS: nonelive in the transcript); removed per review, post-hoc scorer is the authority.doctorintentional-decoupling comment hardened (do not add ledger to the green condition).record_seam_test.pygated;hooks-e2eruns with a hermeticHOMEso "unwired" scenarios don't read the dev's real global config.add-infra-seam-taxonomychange (ratifies the deployment-surface-seam rule; validates--strict).Verified: full smoke green (8 hook self-tests, hooks-e2e 22, mcp/conformance/fault-catch); fresh N-lane acceptance; interactive-path enforcement selftest; two independent cold-read review rounds (findings resolved). No change to the frozen 1.0 contract.
🤖 Generated with Claude Code