Jar ecosystem hardening#2
Merged
Merged
Conversation
…ch, add ecosystem map Deep cross-skill audit of all 23 skills. Applied the highest-leverage fixes; filed the rest to agent-state for follow-up cycles. Structural gate stays green. Fixes: - Align bundled subagent-template names in the reaper/backfill/drift kits with the manifest/install role names (dead-code-reaper-*, test-backfill-*, arch-drift-watcher). The unprefixed names had drifted from the SKILL.md install lines and broke the agents/README prefixed-naming policy; they live inside fenced blocks so the audit gate never saw them. - sprint-ticket-runner: add an explicit launch gate (offers launch, never auto-launches) plus a stop condition. It was the only loop skill that would auto-launch code-writing makers with no termination clause. - Normalize the lone ../references/state-templates.md link to ./ to match its siblings (same resolved target). Docs and state: - Add docs/ecosystem-map.md: intent routing, the two pipeline backbones, the autonomy ladder and human gate, a bundled-vs-external dependency matrix, the 23-skill relationship table, the shared-state map, and the gates note. - File remaining findings to triage-inbox.md (F-1..F-12), audit-policy proposals to decisions.md (HD-1..HD-5), and record completed work plus open tasks in completed.md and loop-state.md. Gate: python scripts/audit-jar.py -> 208 checks, 0 failed. Verified by a separate checker (maker != checker).
The new docs/ecosystem-map.md was unreachable from any entry point. Point the "For agents" paragraph at it so a programmatic reader finds cross-skill routing, the pipeline backbones, and the autonomy ladder next to skills.json.
…delete-safety) Autonomous jar-audit cycle over the four filed Open Tasks; each fixed by a maker and verified by a separate checker (maker != checker). - instrument-observability: add a "When NOT to use" boundary (diagnose-loop / optimization-loop / host bugfix) plus a description NOT-for clause, and a handoff noting its telemetry feeds production-readiness's launch gate. - autonomous-advisor + clean-room: reframe MemBerry / memberry-setup as an optional persistence adapter (clean skip on absence) instead of a hard halt, matching optimization-loop; fix duplicate list numbering. - improve-architecture + dead-code-reaper: name arch-drift-watch as the upstream detector; test-backfill-loop: name agent-state/BUG_TRACKER.md as the canonical suspected-bug sink. - plan-prune: a planning doc may be deleted only once git already holds it; untracked or dirty docs are archived or blocked instead. skills.json regenerated for the instrument-observability description change. Remaining findings F-4..F-11 stay in triage-inbox; completed.md and loop-state.md updated. A checker rejected one inaccurate cross-reference in clean-room mid-cycle; it was corrected and re-verified. Gate: python scripts/audit-jar.py -> 208 checks, 0 failed.
…ap MemBerry row) Hunter swept the surfaces changed this effort plus cross-file consistency and filed one LOW defect: the docs/ecosystem-map.md §4 MemBerry dependency row still pointed at "open findings" (F-1/F-2/F-12) that were closed earlier this cycle. Fixer changed the cell to "optimization-loop, autonomous-advisor, clean-room (all optional)". A separate Validator confirmed the symptom is gone, the new cell matches the implemented optional posture in both skills, no other stale pointer references a closed finding (the F-5 "external" note is correctly still open), scope is one table cell, and the gate is green. BUG-001 -> verified. Gate: python scripts/audit-jar.py -> 208 checks, 0 failed.
…SF-023) Batch 1 of the forge queue (forger != judge throughout). - SF-005 clean-room: 3 independent judges ran the firewall/parity-mode pressure scenario against the patched skill and all returned COMPLY -- the 8 captured rationalizations are refused and the reclassify-to-Transparent escape is closed. SF-005 -> forged (3/3). - SF-023 instrument-observability: a forger applied the GREEN patch closing the captured RED rationalizations (non-waivable investigation gate; high-cardinality identifier governance across tags/extra/context/span; logger-not-a-substitute for the sensitive-surface map; full smoke checklist; an 8-row pressure table) in a 45/+2- diff with the frontmatter description unchanged; 3 independent judges then returned COMPLY. SF-023 -> forged (3/3). Tracker, run packages, completed.md and loop-state.md updated. Forge queue: 6 of 23 forged; SF-006..022 + SF-021 remain pending-red (multi-batch). Gate: python scripts/audit-jar.py -> 208 checks, 0 failed.
Ran the rest of the pending-red queue (SF-006..022, SF-021) as one concurrent RED -> GREEN -> judge x3 pipeline per skill (forger != judge, disjoint files). Every skill surfaced a real shortcut under pressure, was patched to refuse the named dodges (a "Known pressure rationalizations" table plus hard-rule tightening), and passed 3 independent judges. Frontmatter descriptions were not touched, so skills.json stays in sync. Spot-checked the add-to-jar and production-readiness diffs -- sane, on-topic rule tightening, no scope creep. - 17/17 forged 3/3; 0 loopholes, 0 needs-stronger-scenario. - Per-skill run packages written under agent-state/skill-forge-runs/. - Tracker rows set to forged and the queue table de-fragmented (SF-022/023 were orphaned below the rules prose). Completes the forge queue (23/23 forged) and the authorized "all three loops until done" rotation (jar-audit + bug-pipeline + skill-forge). Gate: python scripts/audit-jar.py -> 208 checks, 0 failed.
|
Caution Review failedThe pull request is closed. ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (52)
📝 WalkthroughWalkthroughThis PR formalizes the skill jar's ecosystem knowledge, completes a large skill forge cycle (SF-005..023 → 23/23 forged), and systematically hardens skill documentation across 20+ development and systems-design skills by adding pressure-rationalization tables, stricter gates, and MemBerry optionality clarifications. ChangesEcosystem Knowledge, Audit Closure, and Forge Tracking
Subagent Template Name Standardization
Skill Documentation Hardening and Gate Enforcement
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Poem
✨ Finishing Touches🧪 Generate unit tests (beta)
|
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.
Summary by CodeRabbit
New Features
Documentation
Chores