eea36b9f92
Skills are opt-in (I have to remember to invoke them). Hooks are
ambient — they fire on every matching event automatically. This PR
moves the careful-mode and learnings discipline from "doc I should
read" to "harness-enforced behavior I cannot bypass".
## 6 new hooks (.claude/hooks/)
- pre-bash-careful — REFUSES git push --force to main, rm -rf at root,
DROP TABLE against prod schema. WARNs on force-with-lease, gh pr/
issue close. Tested: blocks the destructive case, allows safe ones.
- pre-edit-freeze — implements /freeze. When .claude/freeze contains
a path glob, edits outside it are denied. Tested: edits to PLAN.md
blocked when scope locked to platform/internal/handlers/.
- session-start-context — auto-loads last 20 cron-learnings, freeze
status, open-PR/issue counts as additionalContext at session start.
Tested: emits valid SessionStart JSON.
- post-edit-audit — appends every Edit/Write to .claude/audit.jsonl
(gitignored). One-line records {ts, tool, file, ok}. Tested writes.
- user-prompt-tag — injects context warnings when prompt mentions
force-push, drop-table, "delete all", "push to main", etc. Tested:
emits warning for "force push the fix to main".
- subagent-stop-judge — off by default; touch .claude/judge-subagents
to enable. When on, prompts orchestrator to verify subagent's last
message addresses the original task. Cost-free MVP (no LLM call yet).
All hooks are Python (jq isn't on the hook PATH on macOS — Python is).
Shared helpers in _lib.py: read_input, deny_pretooluse, add_context,
warn_to_stderr.
## settings.json — wires all 6 hooks
Adds SessionStart, UserPromptSubmit, SubagentStop event handlers.
Existing PreToolUse:Bash + PostToolUse:Edit chains gain the new hooks
alongside the existing ones (check-inbox.sh, echo reminder).
Adds @modelcontextprotocol/server-sequential-thinking MCP server for
structured chain-of-thought scratchpad — useful when triaging multiple
PRs in parallel without losing context.
## .claude/commands/triage.md — slash command shortcut
Manual /triage runs the same flow as the c5074cd5 hourly cron, on
demand. Saves ~4KB of prompt every invocation by pulling the cron
prompt out of working memory.
## CLAUDE.md additions
New "Agent operating rules (auto-loaded — read first)" section right
after Ecosystem Context. Documents:
- Cron / triage discipline (read learnings, treat docs PRs touching
CLAUDE.md/PLAN.md as noteworthy, write per-tick reflections)
- Table of all 6 hooks active in this repo
- List of skills and how to invoke them
- Standing rules (inviolable) consolidated for the agent
This block auto-loads into every conversation context — free behavior
change without me remembering to opt in.
## .gitignore
audit.jsonl, freeze, judge-subagents, per-tick-reflections.md are all
local operational state, never committed.
## Verification
- echo '{"tool_input":{"command":"git push --force origin main"}}' |
bash pre-bash-careful.sh → emits deny JSON ✓
- Same for git status (safe command) → empty output, exit 0 ✓
- pre-edit-freeze with .claude/freeze=platform/handlers/ blocks
edits to PLAN.md, allows edits inside the locked path ✓
- post-edit-audit appends valid JSONL ✓
- session-start-context emits additionalContext with PR/issue counts ✓
- user-prompt-tag emits warning for "force push to main" prompt ✓
- python3 -c "json.load(open('.claude/settings.json'))" → valid ✓
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
65 lines
2.8 KiB
Markdown
65 lines
2.8 KiB
Markdown
---
|
|
name: triage
|
|
description: Run the hourly PR-triage + issue-pickup + code-review + docs-sync loop. Equivalent to one tick of the c5074cd5 cron, on demand.
|
|
---
|
|
|
|
# /triage
|
|
|
|
Manual invocation of the same prompt the hourly cron runs at :17 past each hour. Use when:
|
|
- You want to clear backlog faster than the hourly cadence
|
|
- You're testing a change to the cron prompt itself
|
|
- The cron is session-only and the session has ended
|
|
|
|
## Steps
|
|
|
|
Run the full c5074cd5 cron flow:
|
|
|
|
### Step 0 — Activate guards + replay learnings
|
|
1. Invoke `Skill careful-mode` — load REFUSE/WARN/ALLOW lists.
|
|
2. Read last 20 lines of `~/.claude/projects/-Users-hongming-Documents-GitHub-molecule-monorepo/memory/cron-learnings.jsonl`.
|
|
|
|
### Step 1 — List
|
|
```
|
|
gh pr list --repo Molecule-AI/molecule-monorepo --state open --json number,title,author,isDraft,mergeable,statusCheckRollup,files
|
|
gh issue list --repo Molecule-AI/molecule-monorepo --state open --json number,title,assignees,labels,body
|
|
```
|
|
|
|
### Step 2 — 7-gate verification per PR
|
|
- Gate 1 CI · Gate 2 build · Gate 3 tests · Gate 4 security · Gate 5 design · Gate 6 line review · Gate 7 Playwright if canvas
|
|
- Supplement A: `Skill code-review` on every PR
|
|
- Supplement B: `Skill cross-vendor-review` on noteworthy PRs (auth/billing/data-deletion/migration/large-blast-radius)
|
|
|
|
### Step 2a — Mechanical fixes only
|
|
Fix on-branch + commit `fix(gate-N): ...` + push + poll CI. NEVER fix logic / design / auth issues.
|
|
|
|
### Step 2b — Merge
|
|
All gates pass + 0 🔴 from code-review + cross-vendor agreement → `gh pr merge N --merge --delete-branch`. Merge-commit only.
|
|
|
|
### Step 3 — Docs sync after any merge
|
|
`Skill update-docs` — measure test counts, don't guess. Open `docs/sync-YYYY-MM-DD-tick-N` PR, don't merge.
|
|
|
|
### Step 4 — Issue pickup (cap 2 per tick)
|
|
For each candidate issue: gates I-1..I-6, self-assign, branch, implement, draft PR, run `Skill llm-judge` against issue body + PR diff, mark ready only if score >= 4.
|
|
|
|
### Step 5 — Status report + cron-learnings
|
|
Report includes every subsection (use "none" if empty):
|
|
- Merged: #A, #B
|
|
- Fixed + merged: #C (gate-N fix)
|
|
- Fixed + awaiting CI: #D
|
|
- Skipped-design: #E (🔴 finding)
|
|
- Picked up issue #F → draft PR #G (llm-judge: N/5)
|
|
- Skipped issue #H (gate I-2)
|
|
- Code-review summary: total 🔴/🟡/🔵
|
|
- Cross-vendor pass/escalation
|
|
- Docs PR: #K
|
|
- Idle reason if nothing to do
|
|
|
|
THEN: append 1-3 lines to cron-learnings.jsonl. Terse. Concrete next_action only.
|
|
|
|
## Standing rules (inviolable)
|
|
- Never push to main · Merge-commits only · Dark theme only · No native browser dialogs · Delegate through PM · Only PM mounts the repo
|
|
- careful-mode REFUSE list ALWAYS blocks
|
|
- code-review 🔴 ALWAYS blocks merge
|
|
- cross-vendor disagreement on noteworthy PR escalates to CEO
|
|
- llm-judge ≤ 2 blocks marking a draft PR ready
|