Land review-fix lanes + punch list; merge manager-mode master (bd-dm3 et al.) - #109
Merged
Conversation
- templates/CLAUDE.md: rename "Assurance Surface" -> "Regulated Domain Guardrails" with a regulated-domain-only lead sentence - docs/cli-reference.md: rename the "Govern" command-group label -> "Guardrails" (overview table + section heading); no command names changed Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01STWtVAbCs4gj175c4oWXMJ
Keep the "coordinated team of AI specialists" hero, reframe the lead around plan/team/dispatch/checks, add a concise four-goals list, and replace the stale Why table with the pillar-aware one so the published site matches the README's framing. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01STWtVAbCs4gj175c4oWXMJ
- New docs/pillars.md: high-level four-pillar map with a Mermaid flow of an effort moving through the pillars, per-pillar dive-deeper links, and a pillar->docs mapping table - mkdocs.yml: enable Mermaid (superfences custom fence), surface "The Four Pillars" in nav under Home, refresh site_description to the new framing - docs/index.md: link the homepage to the pillars page - docs/agent-roster.md: add "Pillar context" callout (Pillar 2) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01STWtVAbCs4gj175c4oWXMJ
…ernance pages - architecture.md: "How this maps to the four pillars" section + Mermaid visual - high-level-design.md: pillar-context callout + per-component pillar tags - orchestrator-usage / engine-and-runtime: Pillar 1 & 3 callouts - governance-knowledge-and-events: Pillar 4 callout Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01STWtVAbCs4gj175c4oWXMJ
…arn pages - storage-sync-and-pmo: Pillar 3 callout - observe-learn-and-improve: supporting-layer callout (all four pillars); honest learning-maturity caveats preserved Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01STWtVAbCs4gj175c4oWXMJ
… gap) Make "The Four Pillars" a nav section with pillars.md as the index and four deep pages beneath it. Each page documents the vision, what ships today (grounded in the engine/CLI), and the honest gap between them: - plan-with-foresight: classifier fallback, ±50% forecast, rule-based foresight, complexity-assessment limits, plan-review off by default - compose-the-right-team: limited flavor coverage, on-demand creation friction, heuristic mis-routes, context rot mitigated not eliminated - right-agent-right-time: bd-3f80 token-ceiling gap, claude-teams no-resume, agent_synthesis declared-but-not-dispatched - checks-and-balances: souls experimental + bd-1ca2, signing depends on souls, process-level sandbox, classifier API-key dependency (no self-improving claim) Strict MkDocs build passes; pillars index + nav wired to the new pages. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01STWtVAbCs4gj175c4oWXMJ
Extract a shared requires_audit_coverage predicate so RiskStage injects the auditor for every task ValidationStage would block, run the DataClassifier server-side on the headless/forge path and merge risk as max(LLM, keyword, classifier), and map PlanQualityError to structured 422 responses (with SSE failure event) instead of opaque 500s. Repair the test_archetype_decomposition fixture and document the true gate defaults plus BATON_DEV_MODE/BATON_PLANNER_WARN_ONLY/ BATON_TEAMS_BACKEND_STRICT. Closes bd-xjm, bd-3ba, bd-45e, bd-74e, bd-fnn. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VMHTtWVqejqqnSQFGXsnLL
Probe active-task sqlite state through a temp-copy accessor in core/storage so WAL sidecars never land in team-context, pass --no-optional-locks to every doctor git call, exit 1 when any check errors, resolve BATON_TASK_ID ahead of the sqlite pointer, and record degraded probes in check details instead of swallowing them. Closes bd-2c0, bd-c09, bd-mjm, bd-4j0. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VMHTtWVqejqqnSQFGXsnLL
Wrap per-pack loads so one bad pack no longer truncates the registry, rebuild the TF-IDF index in a finally block, guard non-mapping manifests and frontmatter, and widen error handling to UnicodeDecodeError with actionable invalid-doc-frontmatter, invalid-manifest, and unreadable-doc issues. Closes bd-c5g, bd-sa4. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VMHTtWVqejqqnSQFGXsnLL
Replace the hand-rolled frontmatter line scan in audit_agents_for_teammate_safety with the shared parse_frontmatter utility so zero-indent block sequences flag and comment-only keys do not, and install templates/agents/*.md through all three installers so the talent-builder contract points at files that exist. Closes bd-pg5, bd-3cj. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VMHTtWVqejqqnSQFGXsnLL
…cs sync (bd-dm3) Fourteen low-severity findings from the tiered chunk review (7123076..592bc6e), landed as one pass: - tests/conftest.py: autouse _sandbox_home fixture redirects HOME/ USERPROFILE to a tmp dir so IntelligentPlanner's eager ~/.claude/knowledge load can't leak real packs into ~134 tests (L1) - quickstart.py: knowledge registry now loads with project_root (L2) - plan_cmd.py: --explain + --json in no-save mode emits a single JSON payload with an "explanation" key instead of dropping --json (L3) - knowledge doctor: dotted-stem docs resolve via name+".md" fallback (L4), --knowledge-root expanduser (L5), --strict emits missing-root issue and exits 1 for missing explicit roots (L6) - docs/cli-reference.md: new Knowledge Commands section, 13 subcommands verified against argparse (L7) - talent-manager alias claims deleted from agent-authoring.md and agent-roster.md — alias never existed in AgentRegistry (L8) - team_backends.py: report_path is POSIX-relative (L9); step-specific warnings ordered before static claude-teams caveats so the capped dispatch summary keeps actionable notes (L10) - reference inventory count synced to 20 in CLAUDE.md, GEMINI.md, references/CLAUDE.md, README.md (L11) - executions.py: documented + regression-tested the deliberate semantic split for UnknownTeamBackendError at start vs mid-run (L12) - test_plan_quality_golden.py: autouse delenv for the planner gate env trio (L13) - diagnostics_cmd.py: writable-check caveat (os.access metadata only, NTFS ACLs not evaluated) surfaced in check details (L14) Regression tests added for L3/L4/L6/L9/L10/L12. Full-suite A/B against baseline HEAD confirmed zero regressions (175 failures identical at baseline; Windows-env artifacts: bd backend, POSIX modes, sandbox). Closes bd-dm3. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VMHTtWVqejqqnSQFGXsnLL
Both sides claimed schema migration 46: master shipped manager_mode (INTEGER) and this branch shipped plan_diagnostics (TEXT). Master's v46 is published, so plan_diagnostics is renumbered to v47 and SCHEMA_VERSION bumped to 47; _upsert_plan/_load_plan_struct now carry both columns. ab_cmd.py and the test_plan_cmd_terse.py Path.resolve patch take master's cooperative-parser/Path-equality versions (the shared knowledge __init__.py supports both registration APIs). Post-merge verification: storage, models, cli, api, engine, planning, manager, knowledge suites — 1480 passed; only the 4 known pre-existing environment failures (planner_review, occ_start) remain, identical to the pre-merge baseline. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VMHTtWVqejqqnSQFGXsnLL
DaveGerson
changed the title
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
Lands the full output of the tiered adversarial review of
7123076..592bc6e— 13 fix-lane beads plus the 14-item bd-dm3 punch list — and reconciles the branch with master's manager-mode PMO work. (Supersedes the original README-reframe scope of this PR; those commits are included in the branch history.)Fix lanes (13 beads, P1/P2)
PlanQualityError→ structured 422s (3e67452)baton doctor(temp-copy sqlite probe),BATON_TASK_IDhonored, exit 1 on failed checks (1bf0cfc)finally, frontmatter/Unicode guards) (129bac4)dd7dded)bd-dm3 punch list (14 low-severity items) (
30283b5)Hermetic-test HOME sandbox, quickstart knowledge wiring,
--explain+--jsonJSON payload, dotted-stem doc resolution, expanduser + strict missing-root in knowledge doctor, knowledge CLI docs section, phantom talent-manager alias removed, POSIX report paths, warning ordering, reference-count sync, mid-run backend-error semantics documented+tested, golden-test env isolation, NTFS writability caveat.Merge with master (
0f24f88)Both sides claimed schema migration 46 — master's
manager_modekeeps v46 (published),plan_diagnosticsrenumbered to v47;_upsert_plan/_load_plan_structcarry both columns. ab_cmd and the terse-test Path patch take master's versions.Verification
592bc6e..HEAD: SHIP.🤖 Generated with Claude Code
https://claude.ai/code/session_01VMHTtWVqejqqnSQFGXsnLL