메뉴
Converts a structured decisions document into .claude/rules/ files (frontend.md, backend.md, security.md, infra.md). Called by the rules-writer agent after architect or rules-lock interview completes.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | generate-rules |
| description | Converts a structured decisions document into .claude/rules/ files (frontend.md, backend.md, security.md, infra.md). Called by the rules-writer agent after architect or rules-lock interview completes. |
| argument-hint | <decisions document> |
| user-invocable | false |
You are a rules codifier. You receive a structured decisions document and write it into modular .claude/rules/ files that will be auto-loaded by Claude Code on every session.
Expect a decisions document with these sections (not all required):
Write exactly these files (skip any for which no decisions were made):
.claude/rules/frontend.md# Frontend Rules
## Stack
- [framework, version]
- [component library]
- [styling approach]
## Design System
- [design tokens source]
- [component patterns]
- [animation level and library]
## Conventions
- [naming, file structure, import style]
- [state management approach]
- [accessibility standard]
.claude/rules/backend.md# Backend Rules
## Architecture
- [pattern: MVC / hexagonal / layered / etc.]
- [API style: REST / GraphQL / tRPC / etc.]
- [service boundaries]
## Framework & Runtime
- [framework + version]
- [runtime + version]
## Conventions
- [error handling pattern]
- [logging approach]
- [validation layer]
.claude/rules/security.md# Security Rules
## Authentication
- [strategy: JWT / session / OAuth / etc.]
- [token storage]
- [refresh strategy]
## Rate Limiting
- [implementation: middleware / edge / etc.]
- [limits: N req/min per IP, N req/min per user]
## Input Validation
- [library and where validation happens]
- [sanitization approach]
## Secrets Management
- [where secrets live: env vars / vault / etc.]
- [never hardcode: list of secret types]
## Guardrails
- [CORS policy]
- [CSP headers]
- [other headers]
.claude/rules/infra.md# Infrastructure Rules
## Deployment
- [target: Vercel / AWS / GCP / self-hosted / etc.]
- [environment strategy: staging / prod / etc.]
## Containerization
- [Docker: yes/no, base image]
- [compose: yes/no]
## CI/CD
- [platform: GitHub Actions / GitLab CI / etc.]
- [pipeline stages: lint → test → build → deploy]
## Monitoring
- [error tracking]
- [logging platform]
.claude/rules/. These will be auto-loaded on every Claude Code session."Use when auditing a repository's uncommitted or messy changes and restructuring them into clean, logical commits. Enforces zero AI co-author footers and groups related files by domain.
Systematic codebase analysis — architecture, tech stack, code patterns, brand guidelines, and provider doc health. Produces a structured Codebase Digest consumed by other agents.
Validates and updates provider docs (CLAUDE.md, GEMINI.md, AGENTS.md, WARP.md) for staleness after task execution. Also updates CHANGELOG.md and TASKS.md checklist.
Token-aware session planning. Takes a list of goals/tasks and produces a prioritized TASKS.md checklist split across sessions based on estimated token usage and complexity.
Prompt the user to save the current agent output as a markdown spec file. Handles y/n prompt, target directory selection, filename generation, and file writing. Use at the end of any agent workflow.
Write an Architecture Decision Record (ADR) documenting a technical decision with context, options considered, and rationale. Use when a significant technical decision needs to be documented.