external-auditor
Blind second-opinion audit via subagent. Run with !audit — not automatic.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-10
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- external-auditor
- triggers
- !audit, external audit, blind review, 'second opinion', 'verificá mi auto-score'
- changelog
- docs/ciclos/cycle28-20260815.md
- description
- Blind second-opinion audit via subagent. Run with !audit — not automatic.
- token_budget
- 2525
03 SKILL.md 正文
When to Use
Explicit (!audit), user request, OR required by close-session gate. Auto-required when changes touch: security-scanner/, quality-gate/, auto-metrics/, external-auditor/, immune-system/, ANTI-PATTERN-CATALOG.md, .project.json. Also HIGH-risk (8+ files, auth/storage/API). Skip trivial tasks (~2000 tok waste).
FLOW
- Gather
git diff+ scores | 2. Delegate subagent blind prompt | 3. Timeout/unparseable → retry 1, abort 2nd | 4. Regex-\s*(Correctness|Tokens|ErrPrev|Skill|Speed|Breadth):\s*(\d+); <4 dims → unparseable | 5. Gap =abs(self - audit)| 6. Double-save guard | 7. All gaps ≤1.5 → PASSED · 1 dim >1.5 (self higher) → immune-system · 2+ dims >1.5 → full stop · under-scored → update bias · mixed → immune on over-scored only | 8. Log: bitacora +mem_save(type="audit")| 9. Bias calibration per dim, rolling 3, avg stored.
INTEGRATION
PASSED + avg≥9 + gaps≤1 → mem_save(type="pattern") | OVERSCORE → immune-system | Bitácora: [audit] {date} — {result}: self={s} audit={a} gaps={g}. Gate: close-session.ps1 → needsAudit=true → !audit + PASSED before !close.
BIAS CALIBRATION
Rolling from .learnings/bias-calibration.json:
{ "offsets": { "Correctness": +1.5, ... }, "samples": 3 }
- Bitácora
[audit]today; none → skip, warn. 2. Subtract avg offset per dim BEFORE thresholds. 3. LogBias corrected: {dim}={offset}. 4. <7→immune, ≥9→mem_save. 5. Append (self, audit), rolling 3. No data → OK. Offsets ±>3.0 → flag. Only during!audit/!score.
Anti-Rationalization
| Rationalization | Red Flag | Verification |
|---|---|---|
| "saltarse audit en cambios complejos" | diff toca scanner/gate/metrics sin audit | When to Use: HIGH/protected → audit obligatorio |
| "aceptar output no parseable del auditor" | regex <4 dims sin retry/abort | FLOW §3-4: retry1, abort2nd, regex 4 dims |
| "ignorar gaps de bias calibration" | gap>1.5 sin immune/stop | FLOW §7: gaps≤1.5 PASSED, >1.5 immune/stop |
Red Flags
- Doing work without checking output format → STOP, re-read skill
- Second occurrence of same rationalization → force RED zone
Verification
- Output matches skill ## Output contract + file:line citaton
- cross-ref-check.ps1 → SKILL.md OK
Refs
auto-metrics · immune-system · bitacora · session-resume · quality-gate · security-scanner
Reference
docs/skills/external-auditor/reference.md
04 同仓库更多 Skills
web-quality-audit
Comprehensive web audit: performance, a11y, SEO, responsive, animation, design tokens.
ui-engine
UI system — Grid/Flexbox/@layer/:has(), container queries, compositor-only animation, OKLCH tokens, component patterns
baseline-ui
Anti-slop UI — layout, typography, responsive, animation, tokens. Use for cleanup or polish.
skill-creator
Create new AI agent skills following Agent Skills spec -- auto-bootstrap unknown tasks, define structure, register in AGENTS.md
senior-engineer
Apply senior engineering competencies — system design, trade-off analysis, delegation, mentoring, and tech debt strategy
prompt-engineering
Design effective prompts using SPEARS framework — scope, principles, examples, assertions, refinements, with ReAct and multi-agent patterns