Borrowing it
Nothing to install: this file belongs to cookyman74/my_harness. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/cookyman74/my_harness/main/.claude/skills/external-review-loop/SKILL.mdgit clone --depth 1 https://github.com/cookyman74/my_harnessWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/cookyman74/my_harness/external-review-loop)<a href="https://agentmods.dev/skills/cookyman74/my_harness/external-review-loop"><img src="https://agentmods.dev/badge/skills/cookyman74/my_harness/external-review-loop/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/cookyman74/my_harness/external-review-loop"><img src="https://agentmods.dev/badge/skills/cookyman74/my_harness/external-review-loop.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Excessive Agency · line 50 Skill selects an external model or provider that may use a different account or billing plan than the operator expects. Undisclosed model switches can cause unexpected cost or quota consumption.Fix: Remove the model/provider override or disclose it prominently and require explicit operator approval before invoking an external coding CLI or billed model.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00268 | $0.02425 |
| Opus 5 | $0.00134 | $0.01213 |
| Sonnet 5 | $0.00054 | $0.00485 |
| Haiku 4.5 | $0.00027 | $0.00243 |
Grade A, and why
external-review-loop scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 102 lines — stays where its author put it; the contents beside it link to each section on GitHub.
외부 리뷰 루프 (External Review Loop)
각 단계 산출물마다 codex·gemini에게 리뷰를 요청 → 이슈를 직접 검증 → 확인분만 수정하는 마감 게이트. 일반화 정본 — 어떤 프로젝트·어떤 산출물 유형(설계서/코드/문서)에도 적용.
왜 외부 리뷰인가: 내부 생성-검증/QA는 같은 세션·같은 컨텍스트라 동일한 맹점을 공유한다. 외부 독립 AI는 다른 관점으로 결함을 잡는다.
왜 전건 직접 판정인가: 외부 리뷰어는 설계 결정·동결 계약·기수정 이력·실측 수치를 모른다. 보고 이슈의 상당수(선례 30~40%)는 기각 대상 — 무비판 반영은 계약 파괴·과설계·재작업을 만든다. 판정 권위 = 오케스트레이터(실코드/실문서 대조). 위임 금지.
트리거
- 사용: "외부 리뷰", "codex/gemini 리뷰", "리뷰 게이트", "설계서 리뷰", "코드 리뷰해서 검증·수정", "이슈 검증하고 수정", "단계 산출물 리뷰" — 단계 산출물(설계서/코드/문서)의 마감 게이트로 요청될 때. 사용자가 이슈 목록을 직접 제출하는 수동 리뷰도 포함(Step 4~7 적용).
- 비사용(near-miss): 같은 세션의 내부 자체 검토·생성-검증·QA(독립 관점이 아니므로 별개 게이트). 외부 리뷰어 호출 없이 바로 고치는 단순 수정. 판정·마감 게이트가 필요 없는 단발 질의. 리뷰 없이 코드만 훑어보는 일반 코드 읽기.
입력 (호출 시 채울 플레이스홀더)
{산출물}: 리뷰 대상 경로 — 설계서/코드 디렉토리/문서 (예:docs/design/auth.md또는 변경 파일 목록){단계ID}: 임의 단계 식별자 (예:design-auth,P2-1,feat-login){커밋id}: 해당되면git rev-parse HEAD, 아니면 생략(설계서 리뷰 등){게이트명령}: 프로젝트 테스트/린트 게이트 (예:make gates/npm test && tsc --noEmit/ 없으면 생략)
절차
Step 1 — 리뷰 요청 프롬프트 구성
2종 분담: codex = 일반/정합성 리뷰, gemini = 성능·안정성 리뷰. 산출물 유형에 맞게 "소스코드" → "설계서/문서"로 치환.
# 일반 (codex)
리뷰 대상 : {산출물}
관련 commit id : {커밋id} # 없으면 생략
위 산출물과 관련 자료를 리뷰·검토하여 발생 가능한 이슈를 모두 찾아 보고해줘.
<이슈 작성 방법>
1. [{이슈레벨}] {이슈 타이틀}
- 현황: {현황/상황}
- 이슈: {상세}
- 권고: {해결·대응 방안}
</이슈 작성 방법>
gemini 프롬프트는 동일 틀 + "성능/속도·안정성 중심으로" 한 줄 추가.
Step 2 — 병렬 비대화 실행
프로젝트 루트에서 두 CLI를 백그라운드 병렬·읽기전용 실행. 프롬프트·출력 모두 _workspace/reviews/에 보존(감사 추적 — /tmp 금지).
mkdir -p _workspace/reviews
# 주의: codex exec는 stdin 열려 있으면 "Reading additional input from stdin..." 무한 대기
# → 반드시 < /dev/null 로 stdin 폐쇄
codex exec --sandbox read-only "$(cat _workspace/reviews/{단계ID}_prompt_general.md)" < /dev/null \
> _workspace/reviews/{단계ID}_codex.md 2>&1 &
gemini -p "$(cat _workspace/reviews/{단계ID}_prompt_perf.md)" < /dev/null \
> _workspace/reviews/{단계ID}_gemini.md 2>&1 &
# gemini 긴 프롬프트 대안: cat prompt.md | gemini -p "위 요청대로 리뷰하고 지정 형식으로 보고하라"
- Bash
run_in_background+ timeout 600s. 완료 통보까지 다른 작업 가능. - 실패/타임아웃 → 1회 재시도 → 재실패 시 해당 도구 누락을 결과서에 명시하고 단일 출처로 진행(루프 차단 금지).
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 102 lines · 268 tokens per session scan A 74a737d8cbc3
external-review-loop is a skill published in the GitHub repository cookyman74/my_harness (80 stars, last pushed 5d ago), licensed Apache-2.0. It adds 268 tokens to every session and 2,425 once invoked, about $0.0013 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
autoreview
Pre-commit/ship code review: Codex default; optional Claude or Pi.
rework-rate
Measure and interpret PR rework rate — the emerging 5th DORA metric.
omh-code-review
This is a Hermes-native code-review workflow skill.
revdiff-plan
Review the last Codex assistant message (plan, analysis, or proposal) with inline annotations in a TUI overlay. Extracts the most recent response from Codex rollout files and opens it in revdiff for review and annotation. Activates on "revdiff-plan", "review plan with revdiff", "annotate plan", "review last response"…
code-reviewer
Code review specialist focused on patterns, bugs, security, and performance.
agent-teams-simplify-and-harden
Implementation + audit loop using parallel agent teams with structured simplify, harden, and document passes. Spawns implementation agents to do the work, then audit agents to find complexity, security gaps, and spec deviations, then loops until code compiles cleanly, all tests pass, and auditors find zero issues or…