메뉴
Run a panel of 5 specialized verifiers (correctness, edge-case, security, regression, style) in parallel against a change, with confidence-weighted voting. Disagreement (40-60% pass share) escalates to lead. Replaces single-critic for high-stakes review. Triggers on /aspect-panel, "panel review", "multi-aspect verify", "review with multiple critics".
Run an agent N times in parallel against the same prompt, then aggregate via one of 5 modes — critic argmax, USC consistency, confidence-weighted, hybrid, or AggAgent synthesis. Use for high-stakes invocations where you'd rather pay Nx to be sure. Triggers on /best-of-n, "best of n", "run multiple", "give me three options", "synthesize across rollouts".
Run a complete SCRUM sprint with all ceremonies — Product Owner authors stories, Tech Lead designs, Scrum Master orchestrates implementers, Verifier+Aspect-Panel guard quality, Sprint Auditor adversarially audits, Retro feeds back to /tune-agent. Triggers on /scrum, "run a sprint", "scrum me this", "ship this with full process".
Run an agent N times in parallel against a code-change task, with each rollout in its own sandboxed copy of the codebase, scored by ACTUAL test pass rate (not just critic opinion). Use for SWE-bench-style code changes where the test suite is the ground truth. Triggers on /exec-grounded, "execution-grounded", "verify by running tests", "best-of-N with tests".
Mine past Claude Code session transcripts (JSONL) to extract user corrections, successful patterns, and recurring failure modes. Proposes diffs against lessons.md and rules/*.md for human review. Triggers on /mine-transcripts, "mine my sessions", "what did I learn this week", "session retro". Use weekly or after a fleet completes.
Verify UI changes by capturing before/after screenshots and asking Claude vision to judge whether the change matches intent and didn't break anything else. Mano-verify pattern (arXiv 2509.17336 —
A/B test an agent's current prompt against a candidate variant from policy/<agent>/candidates/. Runs both on the same scenarios, aggregates rewards, recommends promotion if candidate beats current by >1 stderr with n≥10. Triggers on /ab-test, "compare prompts", "test the candidate", "is the new prompt better".
| name | aspect-panel |
| description | Run a panel of 5 specialized verifiers (correctness, edge-case, security, regression, style) in parallel against a change, with confidence-weighted voting. Disagreement (40-60% pass share) escalates to lead. Replaces single-critic for high-stakes review. Triggers on /aspect-panel, "panel review", "multi-aspect verify", "review with multiple critics". |
| when_to_use | Before merging release-blocking changes. Before committing security-sensitive code. After /verify says PASS but you want orthogonal coverage. NOT for trivial diffs (use /verify alone). |
| allowed-tools | Bash, Read |
| arguments | ["target"] |
/aspect-panel <target> spawns 5 aspect-specialized verifiers in parallel, each examining one dimension of the change. Confidence-weighted vote produces a final verdict; disagreement is preserved as a signal.
Built on:
python3 ~/.claude/rl/aspect_panel.py --target "<files-or-diff-or-description>"
The script:
claude -p:
VERDICT: pass|fail, CONFIDENCE: 0.0-1.0, RATIONALE, and RESULT_<aspect>-verifier=PASS|FAIL.pass_share = sum(conf where pass) / sum(conf)pass_share > 0.6 → PASSpass_share < 0.4 → FAIL0.4 ≤ pass_share ≤ 0.6 → ESCALATE (real disagreement; surface to lead){
"ts": "...",
"target": "...",
"aspects": [
{"aspect": "correctness", "verdict": "pass", "confidence": 0.85, "rationale": "..."},
{"aspect": "edge-case", "verdict": "fail", "confidence": 0.72, "rationale": "no handling for None input at src/x.py:42"},
...
],
"aggregate": {
"verdict": "ESCALATE",
"pass_share": 0.55,
"split": true
}
}
A reward event is emitted to ~/.claude/rl/rewards.jsonl recording the panel's decision.
| Stake | Use |
|---|---|
| Trivial change (rename, format, comment) | nothing, or /verify |
| Routine task closure | /verify (single verifier) |
| Pre-commit on a touched-shared-code change | /verify + /aspect-panel |
| Release-blocking, security-sensitive, migration | /aspect-panel + /best-of-n verifier --n 3 |
The panel costs ~5× a single /verify (5 parallel verifiers). Don't run for free.
ChatEval (2308.07201) found homogeneous personas hurt — running 5 identical critics adds noise without signal. Each verifier in our panel has a deliberately different priority: paranoid security vs perf vs maintainability. They catch different bug classes.
The ESCALATE state when they disagree is itself the signal: when 50/50 split, you don't have consensus, and pretending to is worse than admitting it.