agentic-rig

Rig up your project for agentic coding. Analyzes any codebase and generates a complete Claude Code configuration (CLAUDE.md, hooks, skills, subagents, MCP servers).

Install

npm i -g agentic-rig

Or use without installing:

npx agentic-rig install

Usage

agentic-rig install              # Copy skill files to .claude/skills/
agentic-rig uninstall            # Remove installed skill files
agentic-rig status               # Show installation status
agentic-rig init <template>      # Generate config from a project-type template
agentic-rig discover [query]     # Search for community skills
agentic-rig --version            # Print version
agentic-rig --help               # Print usage

Install options

agentic-rig install --force      # Overwrite existing installation without prompting

Init options

agentic-rig init --list          # List available templates
agentic-rig init shopify-theme   # Generate config from the shopify-theme template
agentic-rig init shopify-theme --dry-run   # Preview without writing files
agentic-rig init shopify-theme --force     # Overwrite existing files without prompting
agentic-rig init shopify-theme --dir ./my-project  # Target a specific directory

Discover options

agentic-rig discover react       # Search for React-related community skills
agentic-rig discover testing     # Search for testing skills
agentic-rig discover shopify     # Search for Shopify skills

What it does

/project-setup skill

agentic-rig install copies bundled skills into .claude/skills/ (project-local). This includes:

• project-setup — the main analysis and configuration skill
• find-skills — a meta-skill for searching community skills

Once installed, use /project-setup inside Claude Code to analyze your codebase and generate tailored configuration.

The skill analyzes your project in three phases:

1. Analyze -- detects languages, frameworks, formatters, test tools, and discovers community skills
2. Present plan -- shows a structured recommendation report for your approval, including external skill recommendations (E# items)
3. Generate -- writes CLAUDE.md, hooks, skills, agents, MCP server configs, and installs approved community skills

When a project matches a known template (e.g., Shopify theme, FastAPI app), the skill automatically incorporates domain-specific recommendations marked with [T] prefixes in the report.

/find-skills skill

Use /find-skills <query> inside Claude Code to search the community skill registry at skills.sh. Results include install commands so you can add skills directly.

agentic-rig init

For quick setup without full analysis, agentic-rig init <template> generates a complete Claude Code configuration from a curated project-type template. This writes CLAUDE.md, hooks, skills, agents, and MCP server configs in one step.

After generation, init displays recommended community skills for the template with install commands.

agentic-rig discover

Search the community skill registry from the command line. Runs npx skills find <query> and displays results with install instructions.

agentic-rig discover react

Secure by Default

agentic-rig generates guardrails, not just productivity config. Security is built into every layer.

Guard hooks — PreToolUse hooks block edits to .env files (secrets), lock files (dependency integrity), migration directories (tool-generated), and build artifacts (compiled output). The AI agent physically cannot modify these files.

Security-aware analysis — The analysis phase detects auth, payment, secrets, and user-data patterns. When found, generated configs include security-specific CLAUDE.md guidance and a security-reviewer subagent.

Domain-specific security knowledge — Templates encode framework-level security practices: webhook signature validation (Shopify), output escaping (Liquid), CORS configuration (FastAPI), server-side token handling (Next.js).

Continuous improvement — The feedback loop tracks which guardrails users keep or skip. self-improve uses this data to evolve security recommendations based on real usage.

Human-in-the-loop — every recommendation is presented for review; nothing is applied silently. Use --dry-run to inspect everything before any files are written.

agentic-rig is not a replacement for security audits, penetration testing, or professional security review. It generates sensible defaults that reduce the surface area for common mistakes.

Available Templates

| Template | Description | |----------|-------------| | shopify-theme | Shopify theme development with Liquid, Dawn, and Online Store 2.0 | | shopify-app | Shopify app development with React Router, Polaris, Prisma, and extensions | | nextjs-sanity | Next.js with Sanity CMS, App Router, GROQ, and Sanity Studio | | python-fastapi | Python API development with FastAPI, Pydantic, and SQLAlchemy |

Run agentic-rig init --list to see all available templates.

Community Skills

Each template includes curated external skill recommendations from the skills.sh ecosystem. These are shown after agentic-rig init and as E# items during /project-setup analysis.

You can also discover skills manually:

# Via the CLI
agentic-rig discover react

# Via npx directly
npx skills find react

# Install a skill
npx skills add vercel-labs/agent-skills --skill vercel-react-best-practices

Browse all available skills at skills.sh.

Requirements

• Node.js >= 18
• Claude Code

License

MIT