Jar ecosystem hardening

README.md

@@ -149,7 +149,7 @@ the generated index, Skill Forge, and the repo's hook evidence flow.
 
 ## For agents
 
-Reading this repo programmatically? Route from [`skills.json`](skills.json) — a generated index of every skill's `name`, routing `description`, `path`, `tags`, `core`, and optional maturity/evidence fields (one fetch, no directory crawl; it's gate-checked against the frontmatter, so it can't drift). Install via the plugin marketplace above (Claude Code) or copy a skill's folder (any host). Every push is verified by `python scripts/audit-jar.py` — the badge above is that gate. If you *operate* in this repo (run a loop cycle, fix a bug), the rules in [`AGENTS.md`](AGENTS.md) bind you.
+Reading this repo programmatically? Route from [`skills.json`](skills.json) — a generated index of every skill's `name`, routing `description`, `path`, `tags`, `core`, and optional maturity/evidence fields (one fetch, no directory crawl; it's gate-checked against the frontmatter, so it can't drift). For how the skills **compose and route between each other** — the pipeline backbones, the autonomy ladder, and which skill defers to which — read [`docs/ecosystem-map.md`](docs/ecosystem-map.md). Install via the plugin marketplace above (Claude Code) or copy a skill's folder (any host). Every push is verified by `python scripts/audit-jar.py` — the badge above is that gate. If you *operate* in this repo (run a loop cycle, fix a bug), the rules in [`AGENTS.md`](AGENTS.md) bind you.
 
 Development and systems-design skills also ship generated sub-agent packs in [`development/agents/`](development/agents/README.md) and [`systems-design/agents/`](systems-design/agents/README.md). Each pack's `manifest.json` is the source of truth; `python scripts/gen-agent-packs.py` renders copy-ready Claude Code and Codex agent files, and the audit gate verifies they stay in sync. Install only the roles a loop, panel, or design review actually needs.
 

agent-state/BUG_TRACKER.md

@@ -4,22 +4,35 @@
 
 | Field | Value |
 |---|---|
-| Last Hunter Scan | 2026-06-09T18:30:00Z |
-| Last Fixer Pass | |
-| Last Validator Pass | |
-| Total Found | 0 |
+| Last Hunter Scan | 2026-06-12T00:00:00Z |
+| Last Fixer Pass | 2026-06-12T00:00:00Z |
+| Last Validator Pass | 2026-06-12T00:00:00Z |
+| Total Found | 1 |
 | Total Pending | 0 |
 | Total In Progress | 0 |
 | Total Fixed | 0 |
 | Total In Validation | 0 |
-| Total Verified | 0 |
+| Total Verified | 1 |
 | Total Reopened | 0 |
 | Total Blocked | 0 |
 
 ---
 
 ## Sweep Notes
 
+**Sweep 2 — 2026-06-12T00:00:00Z**
+Focus: 8 edited skills (instrument-observability, autonomous-advisor, clean-room, dead-code-reaper, improve-architecture, plan-prune, test-backfill-loop, sprint-ticket-runner); 3 edited kits (reaper-kit.md, backfill-kit.md, drift-kit.md); loop-architecture.md; docs/ecosystem-map.md; cross-file consistency against development/agents/manifest.json and systems-design/agents/manifest.json.
+
+1 finding filed (BUG-001). Verified clean:
+- Kit template `name:` fields in all three kits match manifest role names exactly (dead-code-reaper-scout/-reaper/-validator, test-backfill-scout/-writer/-verifier, arch-drift-watcher).
+- Ecosystem-map §5 installable roles column is consistent with manifest for all checked rows.
+- All `references/` links in instrument-observability (investigation-model.md, instrumentation-playbook.md, sentry-patterns.md) resolve.
+- MemBerry is correctly marked optional (clean skip) in both autonomous-advisor:57 and clean-room:262.
+- plan-prune delete precondition (committed-clean only) is in place at SKILL.md:51.
+- sprint-ticket-runner Operating Contract carries the launch gate and stop condition.
+- loop-architecture.md companion links (state-templates.md, subagent-templates.md, automation-templates.md, safety-and-gates.md, worktree-isolation.md, role-skills/) and cross-skill link to optimization-loop all resolve.
+- Defect: ecosystem-map:139 MemBerry row retains stale "see open findings" pointer for autonomous-advisor & clean-room after F-1/F-2/F-12 were closed by jar-audit-eco-1 (filed as BUG-001).
+
 **Sweep 1 — 2026-06-09T18:30:00Z**
 Focus: `scripts/audit-jar.py` logic bugs, `loop-engineering/scripts/scaffold-loop.py` logic bugs, cross-file consistency (`agent-state/loop-state.md`, `docs/prompts/jar-audit-driver.md`, `docs/prompts/bug-pipeline-driver.md`, `AGENTS.md`).
 
@@ -32,4 +45,34 @@ No findings above the bar. All code paths traced and verified correct:
 - Cross-file path references consistent across `AGENTS.md`, `loop-state.md`, `jar-audit-driver.md`, and `bug-pipeline-driver.md`.
 - Audit gate runs clean: 27 checks, 0 failed.
 
+---
+
+## BUG-001 — Ecosystem-map §4 MemBerry row says "see open findings" after F-1/F-2/F-12 were closed
+
+| Field | Value |
+|---|---|
+| ID | BUG-001 |
+| Status | verified |
+| Severity | low |
+| File | docs/ecosystem-map.md:139 |
+| Filed by | hunter sweep 2026-06-12 |
+
+**Title:** Ecosystem-map §4 MemBerry dependency row still reads "should be optional — see open findings" for autonomous-advisor & clean-room after the open findings (F-1, F-2, F-12) were closed by jar-audit-eco-1.
+
+**Evidence:** `docs/ecosystem-map.md:139`
+```
+| MemBerry + `memberry-setup` | ... | optimization-loop (optional), autonomous-advisor & clean-room (should be optional — see open findings) | optional persistence adapter; absent = files-only |
+```
+`agent-state/triage-inbox.md:11` confirms F-1, F-2, F-12 were "RESOLVED by jar-audit-eco-1". Both `autonomous-advisor/SKILL.md:57` and `clean-room/SKILL.md:262` now say MemBerry is an optional adapter with a clean skip on absence — the "should be optional" concern and the "see open findings" pointer are both obsolete.
+
+**Observable symptom:** A fresh agent reading §4 of the ecosystem-map will see "see open findings" and search the triage-inbox for an actionable open finding about MemBerry optionality in these two skills. No such finding exists (the inbox header explicitly marks them resolved), causing confusion, wasted investigation, or a spurious "fix" attempt that re-edits correctly-implemented code.
+
+**Repro:** Read `docs/ecosystem-map.md:139`; then read `agent-state/triage-inbox.md` header (lines 9–14) — the cross-reference resolves to an already-closed item.
+
+**Fix scope:** Change the "Used by" cell for MemBerry from `autonomous-advisor & clean-room (should be optional — see open findings)` to `autonomous-advisor & clean-room (optional)` — matching the already-implemented posture.
+
+**Fixer (jar-audit-eco-1, 2026-06-12):** Edited `docs/ecosystem-map.md:139` — the MemBerry "Used by" cell now reads `optimization-loop, autonomous-advisor, clean-room (all optional)`; the stale "see open findings" pointer is removed. Smallest diff (one table cell). `python scripts/audit-jar.py` -> 208 checks, 0 failed (ecosystem-map links still resolve). Status -> fixed; awaiting independent validator.
+
+**Validator (independent, 2026-06-12):** VERIFIED. ecosystem-map.md:139 no longer contains "see open findings" or "should be optional"; cell now reads `optimization-loop, autonomous-advisor, clean-room (all optional)`. Implemented posture confirmed optional/clean-skip in autonomous-advisor/SKILL.md:57 ("optional persistence adapter, not a prerequisite ... clean skip, never a halt") and clean-room/SKILL.md:262 ("optional persistence adapter — its absence is a clean skip, not a blocker"). Grep of whole ecosystem-map for stale pointers: only :70 (unrelated prose) and :141 (references F-5, still OPEN in triage-inbox — correctly not stale). `python scripts/audit-jar.py` -> `Summary: 208 checks, 0 failed.` (exit 0). `git diff --stat` -> only docs/ecosystem-map.md (1 row) + agent-state/BUG_TRACKER.md; diff is one table cell, not a rewrite. F-1/F-2/F-12 confirmed closed (triage-inbox header lines 11-14; completed.md C-2026-06-12-T-ECO-2). Status -> verified.
+
 <!-- HUNTER: Append new bugs above this line -->

agent-state/SKILL_FORGE_TRACKER.md

@@ -26,23 +26,25 @@ removed or renamed, mark the row `blocked` and record the decision in
 | SF-002 | auto-research | development | `development/auto-research/SKILL.md` | forged | 3/3 | fixed-budget experiment shortcut pressure | `agent-state/skill-forge-runs/auto-research.md` | complete |
 | SF-003 | autonomous-advisor | development | `development/autonomous-advisor/SKILL.md` | forged | 3/3 | hands-off PRP guardrail pressure | `agent-state/skill-forge-runs/autonomous-advisor.md` | complete |
 | SF-004 | bug-pipeline | development | `development/bug-pipeline/SKILL.md` | forged | 3/3 | hunter/fixer/validator shortcut pressure | `agent-state/skill-forge-runs/bug-pipeline.md` | complete |
-| SF-005 | clean-room | development | `development/clean-room/SKILL.md` | patched | 0/3 | firewall and parity-mode pressure | `agent-state/skill-forge-runs/clean-room.md` | REFACTOR judge run 1 |
-| SF-006 | dead-code-reaper | development | `development/dead-code-reaper/SKILL.md` | pending-red | 0/3 | unsafe deletion pressure | - | RED scenario |
-| SF-007 | design-panel | development | `development/design-panel/SKILL.md` | pending-red | 0/3 | single-design shortcut pressure | - | RED scenario |
-| SF-008 | diagnose-loop | development | `development/diagnose-loop/SKILL.md` | pending-red | 0/3 | premature fix pressure | - | RED scenario |
-| SF-009 | improve-architecture | development | `development/improve-architecture/SKILL.md` | pending-red | 0/3 | shallow refactor pressure | - | RED scenario |
-| SF-010 | loop-engineer | development | `development/loop-engineer/SKILL.md` | pending-red | 0/3 | vague-loop autonomy pressure | - | RED scenario |
-| SF-011 | optimization-loop | development | `development/optimization-loop/SKILL.md` | pending-red | 0/3 | metric and backlog shortcut pressure | - | RED scenario |
-| SF-012 | plan-prune | development | `development/plan-prune/SKILL.md` | pending-red | 0/3 | stale-plan consolidation pressure | - | RED scenario |
-| SF-013 | review-panel | development | `development/review-panel/SKILL.md` | pending-red | 0/3 | unverified finding pressure | - | RED scenario |
-| SF-014 | skill-forge | development | `development/skill-forge/SKILL.md` | pending-red | 0/3 | self-forging rationalization pressure | - | RED scenario |
-| SF-015 | sprint-ticket-runner | development | `development/sprint-ticket-runner/SKILL.md` | pending-red | 0/3 | parallelism and sprint-drift pressure | - | RED scenario |
-| SF-016 | test-backfill-loop | development | `development/test-backfill-loop/SKILL.md` | pending-red | 0/3 | non-biting test pressure | - | RED scenario |
-| SF-017 | api-design | systems-design | `systems-design/api-design/SKILL.md` | pending-red | 0/3 | protocol and idempotency shortcut pressure | - | RED scenario |
-| SF-018 | data-store-selection | systems-design | `systems-design/data-store-selection/SKILL.md` | pending-red | 0/3 | brand-choice and shard-key pressure | - | RED scenario |
-| SF-019 | design-system | systems-design | `systems-design/design-system/SKILL.md` | pending-red | 0/3 | premature complexity pressure | - | RED scenario |
-| SF-020 | production-readiness | systems-design | `systems-design/production-readiness/SKILL.md` | pending-red | 0/3 | launch-without-drill pressure | - | RED scenario |
-| SF-021 | unit-test-quality | development | `development/unit-test-quality/SKILL.md` | pending-red | 0/3 | AI slop tests, weak assertions, and coverage-metric pressure | - | RED scenario |
+| SF-005 | clean-room | development | `development/clean-room/SKILL.md` | forged | 3/3 | firewall and parity-mode pressure | `agent-state/skill-forge-runs/clean-room.md` | complete |
+| SF-006 | dead-code-reaper | development | `development/dead-code-reaper/SKILL.md` | forged | 3/3 | unsafe deletion pressure | `agent-state/skill-forge-runs/dead-code-reaper.md` | complete |
+| SF-007 | design-panel | development | `development/design-panel/SKILL.md` | forged | 3/3 | single-design shortcut pressure | `agent-state/skill-forge-runs/design-panel.md` | complete |
+| SF-008 | diagnose-loop | development | `development/diagnose-loop/SKILL.md` | forged | 3/3 | premature fix pressure | `agent-state/skill-forge-runs/diagnose-loop.md` | complete |
+| SF-009 | improve-architecture | development | `development/improve-architecture/SKILL.md` | forged | 3/3 | shallow refactor pressure | `agent-state/skill-forge-runs/improve-architecture.md` | complete |
+| SF-010 | loop-engineer | development | `development/loop-engineer/SKILL.md` | forged | 3/3 | vague-loop autonomy pressure | `agent-state/skill-forge-runs/loop-engineer.md` | complete |
+| SF-011 | optimization-loop | development | `development/optimization-loop/SKILL.md` | forged | 3/3 | metric and backlog shortcut pressure | `agent-state/skill-forge-runs/optimization-loop.md` | complete |
+| SF-012 | plan-prune | development | `development/plan-prune/SKILL.md` | forged | 3/3 | stale-plan consolidation pressure | `agent-state/skill-forge-runs/plan-prune.md` | complete |
+| SF-013 | review-panel | development | `development/review-panel/SKILL.md` | forged | 3/3 | unverified finding pressure | `agent-state/skill-forge-runs/review-panel.md` | complete |
+| SF-014 | skill-forge | development | `development/skill-forge/SKILL.md` | forged | 3/3 | self-forging rationalization pressure | `agent-state/skill-forge-runs/skill-forge.md` | complete |
+| SF-015 | sprint-ticket-runner | development | `development/sprint-ticket-runner/SKILL.md` | forged | 3/3 | parallelism and sprint-drift pressure | `agent-state/skill-forge-runs/sprint-ticket-runner.md` | complete |
+| SF-016 | test-backfill-loop | development | `development/test-backfill-loop/SKILL.md` | forged | 3/3 | non-biting test pressure | `agent-state/skill-forge-runs/test-backfill-loop.md` | complete |
+| SF-017 | api-design | systems-design | `systems-design/api-design/SKILL.md` | forged | 3/3 | protocol and idempotency shortcut pressure | `agent-state/skill-forge-runs/api-design.md` | complete |
+| SF-018 | data-store-selection | systems-design | `systems-design/data-store-selection/SKILL.md` | forged | 3/3 | brand-choice and shard-key pressure | `agent-state/skill-forge-runs/data-store-selection.md` | complete |
+| SF-019 | design-system | systems-design | `systems-design/design-system/SKILL.md` | forged | 3/3 | premature complexity pressure | `agent-state/skill-forge-runs/design-system.md` | complete |
+| SF-020 | production-readiness | systems-design | `systems-design/production-readiness/SKILL.md` | forged | 3/3 | launch-without-drill pressure | `agent-state/skill-forge-runs/production-readiness.md` | complete |
+| SF-021 | unit-test-quality | development | `development/unit-test-quality/SKILL.md` | forged | 3/3 | AI slop tests, weak assertions, and coverage-metric pressure | `agent-state/skill-forge-runs/unit-test-quality.md` | complete |
+| SF-022 | add-to-jar | development | `development/add-to-jar/SKILL.md` | forged | 3/3 | drop-in skill pressure | `agent-state/skill-forge-runs/add-to-jar.md` | complete |
+| SF-023 | instrument-observability | development | `development/instrument-observability/SKILL.md` | forged | 3/3 | observability shortcut pressure | `agent-state/skill-forge-runs/instrument-observability.md` | complete |
 
 ## Run Package Rules
 
@@ -53,5 +55,3 @@ removed or renamed, mark the row `blocked` and record the decision in
   `python scripts/audit-jar.py` exiting 0.
 - If a public skill contract change is required, mark `blocked` and write the
   decision row before editing.
-| SF-022 | add-to-jar | development | `development/add-to-jar/SKILL.md` | pending-red | 0/3 | drop-in skill pressure | - | RED scenario |
-| SF-023 | instrument-observability | development | `development/instrument-observability/SKILL.md` | red-captured | 0/3 | observability shortcut pressure | `agent-state/skill-forge-runs/instrument-observability.md` | GREEN patch |