agentskills.codes
RP

rp-eng-review

by ryan-papaCodingSource

Runs an automated 5-point engineering review on PRDs to ensure technical feasibility and architecture quality.

Code ReviewArchitectureAgents

Install

mkdir -p .claude/skills/rp-eng-review && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/12874" && unzip -o skill.zip -d .claude/skills/rp-eng-review && rm skill.zip

Installs to .claude/skills/rp-eng-review

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

[5] 엔지 리뷰. Codex-led 5항목 점수제 독립 리뷰. Use when Codex should run or follow the harness rp-eng-review stage, or when the user invokes /rp-eng-review, $rp-eng-review, or asks for this workflow step.
193 chars✓ has a “when” trigger
Intermediate

Key capabilities

  • →Validate technical feasibility of PRDs
  • →Verify architectural alignment of PRDs
  • →Score PRDs using a 5-item harness
  • →Reflect review findings into the PRD body
  • →Manage re-execution of reviews after revisions

How it works

This skill uses subagents to score a PRD against five criteria, then integrates the findings directly into the PRD document.

Inputs & outputs

You give it
A Product Requirements Document (PRD)
You get back
An updated PRD with review findings and a pass/fail determination

When to use rp-eng-review

  • →Validating PRD technical feasibility
  • →Checking architectural consistency
  • →Automating PRD quality scoring
  • →Reviewing project workflow compliance

About this skill

Generated from docs/skills/rp-eng-review.md. Do not edit this file directly; run python3 scripts/sync-codex-skills.py after changing the Claude skill source.

Codex Adapter Notes

  • Treat this as the Codex project-local equivalent of the matching Claude /rp-* skill.
  • Resolve copied relative links against the source file under docs/skills/ when needed.
  • Do not record Claude-only .claude hooks or slash commands as executed unless they actually ran.
  • Use Codex spawn_agent for independent review when a review step requires role separation.
  • Runtime is the reviewer: never call Claude for an extra cross-runtime review.
  • Reflect Codex-Lead review findings into the PRD body itself; do not write review-codex-*.md or review-claude-*.md evidence files.

rp-eng-review

엔지니어링 리뷰. PRD의 기술적 실현 가능성과 아키텍처 적합성 검증. 메인 셀프 채점 절대 금지 — 해당 런타임 서브에이전트가 기술 관점으로 채점.

작성 모드

모드채점 주체외부 추가 리뷰
Claude-Lead (Claude Code)Claude Agent 툴 서브에이전트없음 — Codex 호출 금지
Codex-Lead (Codex CLI)Codex spawn_agent 서브에이전트없음 — Claude 호출 금지

런타임 = 리뷰어. 교차 런타임 추가 리뷰 전면 금지. SSOT: ../harness-absolute-rules.md "작성 모드 및 리뷰 매트릭스".

트리거

  • PRD 작성 완료 후 — [4] 기획 리뷰와 동시 발사 (병렬)
  • /rp-eng-review 명령

절차 (양 모드 공통)

  1. 서브에이전트 병렬 발사 (Claude-Lead=Agent 툴 subagent_type=general-purpose / Codex-Lead=spawn_agent): [4]·[5] 두 서브에이전트를 동일 메시지에서 동시 발사, 컨텍스트 미공유 독립 판정. 메인 셀프 채점 금지. 프롬프트 4 필수 항목 (a)~(d) → SSOT: ../harness-absolute-rules.md "[리뷰 단계 서브에이전트 필수]" 절. 본 단계 적용값: (a) PRD + CLAUDE.md·harness-db.md 참고 (b) 5항목 (c)·(d) SSOT 그대로
    • Codex-Lead 대칭: spawn_agent 동시 발사 불가 환경이면 순차 fallback 허용 — 판정 규칙·통과 조건 동일
  2. 판정: 서브에이전트 결과만으로 판정
  3. 결과 처리: 별도 파일 저장 없음. 점수·지적은 인-메모리에서 PRD 본문에 반영
  4. 통과: 평균 ≥8.0 + 각 항목 ≥7 → [4]도 통과 시 다음 단계
  5. 미달: 지적 반영 → PRD 본문 갱신 → 미달 축만 새 서브에이전트로 재실행. 최대 2회 추가, 총 3회(축별 독립 카운터). 3회 미달 시 자동 중단 + 사용자 결정 요청
  6. 조건부 재검증: PRD 수정이 아키텍처·데이터 모델·성능·보안·운영 서술을 건드리면 통과한 [5]도 1회 재검증. 전면 개작 시 무조건 양축 재실행. 재검증은 카운트 미소비이며 같은 사이클 내 연쇄 재검증 금지 — 재검증이 미달이면 그 축은 다음 사이클의 미달 축으로 전환
  7. 기술 실패 Fallback: SSOT ../harness-absolute-rules.md "재시도 한도" 적용. 한 축만 실패 시 해당 축만 최대 2회 재호출(총 3회), 정상 축 결과 보존. 실패 축을 생략한 통과 판정 금지. 메인 셀프 채점 우회 금지

평가 항목

항목설명
아키텍처기술 스택, 구조 적합성
확장성트래픽/데이터 증가 대응
보안인증, 데이터 보호, 취약점
성능응답 시간, 리소스 효율
운영성배포, 모니터링, 장애 대응

판정

  • 평균 >= 8.0 + 각 항목 >= 7 → 통과
  • 평균 미달 또는 항목별 최저 < 7 → 문서 작성자가 PRD 본문 갱신 후 재실행 (인-메모리 최대 3회)
  • 3회 미달 → 자동 중단 + 사용자 결정 요청 (강행/재설계/중단)

▶ 자동 전환

병렬 축 [4] 기획 리뷰와 양축 통과 시 ✓ [4]·[5] 리뷰 통과 (Claude-Lead) 또는 ✓ [4]·[5] 리뷰 통과 (Codex-Lead) 출력 후 /rp-task 자동 진입. 한 축만 통과면 미달 축 재실행 후 재판정.

→ PRD 상세: ../harness-prd.md

When not to use it

  • →When the main agent should self-score the PRD
  • →When external Codex re-invocation is needed in Claude-led mode
  • →When review findings should be saved to separate evidence files

Limitations

  • →The main agent is prohibited from self-scoring
  • →Review findings are not saved to external files
  • →Re-execution is limited to a maximum of 2 additional attempts

How it compares

This workflow automates the scoring and integration of review feedback into the PRD, unlike a manual review process that requires separate documentation and updates.

Compared to similar skills

rp-eng-review side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
rp-eng-review (this skill)02moNo flagsIntermediate
parallel-agents66moNo flagsIntermediate
multi-agent-brainstorming14moNo flagsIntermediate
architect-review1094moNo flagsAdvanced

Try saying

Example prompts that trigger this skill in your AI assistant.

More by ryan-papa

View all by ryan-papa →

rp-dev

ryan-papa

[7] 개발. 태스크별 반복 사이클. 통합 브랜치에서 태스크 단위 구현·테스트. Use when Codex should run or follow the harness rp-dev stage, or when the user invokes /rp-dev, $rp-dev, or asks for this workflow step.

00

You might also like

parallel-agents

davila7

Multi-agent orchestration patterns. Use when multiple independent tasks can run with different domain expertise or when comprehensive analysis requires multiple perspectives.

634

multi-agent-brainstorming

sickn33

Use this skill when a design or idea requires higher confidence, risk reduction, or formal review. This skill orchestrates a structured, sequential multi-agent design review where each agent has a strict, non-overlapping role. It prevents blind spots, false confidence, and premature convergence.

10

architect-review

sickn33

Master software architect specializing in modern architecture patterns, clean architecture, microservices, event-driven systems, and DDD. Reviews system designs and code changes for architectural integrity, scalability, and maintainability. Use PROACTIVELY for architectural decisions.

109320

mcp-builder

anthropics

Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).

136215

solid-principles

SmidigStorm

Enforce SOLID principles (Single Responsibility, Open/Closed, Liskov Substitution, Interface Segregation, Dependency Inversion) in object-oriented design. Use when writing or reviewing classes and modules.

57236

codex

Lucklyric

Invoke Codex CLI for complex coding tasks requiring high reasoning capabilities. This skill should be invoked when users explicitly mention "Codex", request complex implementation challenges, advanced reasoning, or need high-reasoning model assistance. Automatically triggers on codex-related requests and supports session continuation for iterative development.

32238

Search skills

Search the agent skills registry