{"manifest":{"name":"CLAUDE.md Generator","version":"1.0.0","description":"Teaches agents how to analyze a codebase and generate effective CLAUDE.md files with progressive disclosure.","tags":["meta","context","generator","toolkit"],"standard":"agentskills.io","standard_version":"1.0","content_checksum":"b9b1bab7c3fe2650bd504065983ea28eefaf2d2c7991aae6629ec6e1c88e2b6a","bundle_checksum":null,"metadata":{},"files":[]},"files":{"SKILL.md":"# CLAUDE.md Generator\n\n> **Purpose:** Teach an AI agent how to analyze a codebase and produce a high-quality CLAUDE.md file with progressive disclosure.\n\n---\n\n## Overview\n\nA great CLAUDE.md is the single most impactful file in a repository for AI-assisted development. It tells the agent what it needs to know — and nothing more. This skill teaches you how to create one.\n\n---\n\n## Principles\n\n1. **Concise over comprehensive** — 200 lines max for CLAUDE.md. Details go in agent_docs/.\n2. **Universal, not personal** — Write for any agent, not a specific model.\n3. **No linting rules** — Agents get these from config files. Don't duplicate.\n4. **Progressive disclosure** — Top-level has critical info. agent_docs/ has deep dives.\n5. **Imperative rules** — \"Never do X\" and \"Always do Y\" are clearer than explanations.\n\n---\n\n## Process\n\n### Step 1: Analyze the Codebase\n\nScan these files first:\n- `package.json` / `Cargo.toml` / `pyproject.toml` — stack and dependencies\n- Existing config files (`.eslintrc`, `tsconfig.json`, etc.) — already-encoded rules\n- `README.md` — project purpose (but don't copy it)\n- Recent git history — active patterns and conventions\n\n### Step 2: Identify Critical Rules\n\nLook for:\n- **Data access patterns** — ORMs, query builders, column selection rules\n- **Numeric conventions** — score ranges (0-1 vs 0-100), currency units\n- **Auth patterns** — how auth is handled, typed clients, middleware\n- **Architecture constraints** — \"never import X from Y\", module boundaries\n- **Testing patterns** — test file locations, mocking conventions, assertion styles\n- **Common gotchas** — things that break builds or cause subtle bugs\n\n### Step 3: Structure the CLAUDE.md\n\n```markdown\n# {Project Name}\n\n{One paragraph: what the project is and does}\n\n## Stack\n\n- {Framework} + {Language} + {Styling}\n- {Database/backend}\n- {Testing framework}\n\n## Quick Reference\n\n```bash\n{dev command}      # {what it does}\n{test command}     # {what it does}\n{build command}    # {what it does}\n```\n\n## Workflow\n\n- {Branch strategy}\n- {Deploy process}\n\n## Critical Rules\n\n- **{Rule name}** — {concise explanation}\n- **{Rule name}** — {concise explanation}\n\n## Agent Docs (read when relevant)\n\n| File | When to read |\n|------|-------------|\n| `agent_docs/{file}.md` | {trigger condition} |\n```\n\n### Step 4: Create agent_docs/ Files\n\nFor each deep topic, create a focused file:\n\n- `agent_docs/project_structure.md` — file layout, key directories, naming conventions\n- `agent_docs/development_workflow.md` — git workflow, environments, migrations, deployment\n- `agent_docs/common_gotchas.md` — debugging issues, form validation, auth quirks\n- `agent_docs/testing.md` — writing and running tests, mocking patterns\n\nEach agent_docs file should be **self-contained** — an agent reads it independently.\n\n---\n\n## Anti-Patterns\n\n- Copying the entire README into CLAUDE.md\n- Listing every file in the project\n- Including linting rules the agent gets from config\n- Writing paragraphs when bullet points suffice\n- Including setup instructions (agents don't run npm install)\n- Documenting obvious conventions (standard Next.js patterns, etc.)\n\n---\n\n## Quality Checklist\n\n- [ ] CLAUDE.md is under 200 lines\n- [ ] Every rule is actionable (\"Never X\", \"Always Y\")\n- [ ] No duplicate info from config files\n- [ ] agent_docs/ table has clear \"when to read\" triggers\n- [ ] A new developer (or agent) could be productive after reading just CLAUDE.md\n- [ ] No personal preferences disguised as project rules"}}