@monomit/primer

scaffold AI-ready projects

The CLI for primer — a tool that scaffolds new projects with pre-configured AI agent conventions for Cursor and Claude Code.

Install

npm install -g @monomit/primer
# or
pnpm add -g @monomit/primer

Commands

primer init

Scaffold a new AI-ready project interactively.

primer init           # AI-assisted (generates tailored agent roles)
primer init --offline # static scaffold only, no API call

Prompts for: project name, package manager, AI tools (Cursor / Claude Code), skill packages, git init, AI provider, project description, stack, and constraints.

Generates:

• AGENTS.md — mandatory agent entry point
• GETTING_STARTED.md — exact first steps
• .primer/project.json — persisted config
• Canonical docs (CANONICAL_ARCHITECTURE.md, CANONICAL_PATTERNS.md)
• AI tool configs (.cursor/rules/, .claude/commands/)
• Selected skill packages
• AI-generated agent role docs in docs/agents/

primer retrofit

Add primer conventions to an existing project.

primer retrofit            # interactive, detects existing config
primer retrofit --dry-run  # preview changes without writing
primer retrofit --force    # overwrite existing files

Detects your package manager, AI tools, and existing skills from the project. Only adds what is missing.

primer brief-me

Generate a technical brief from your project documentation.

primer brief-me                    # full project brief → docs/BRIEF.md
primer brief-me --domain database  # scoped brief → docs/briefs/database.md
primer brief-me --offline          # raw doc summary, no API call

Reads docs/CANONICAL_ARCHITECTURE.md, docs/CANONICAL_PATTERNS.md, docs/agents/, and relevant skill knowledge docs. Produces a brief with architecture summary, domain breakdown, key patterns, agent roles, critical constraints, next steps, and an exact "Start Building" sequence.

Valid domains: backend, database, auth, frontend, testing.

Skill packages

Select skill packages during primer init or primer retrofit:

| Skill | Domain | |---|---| | database | Schema design, migrations, performance, security | | auth | Authentication, authorization, secrets, incident response | | backend | API contracts, resilience, traffic control, observability | | frontend | RSC boundary, hydration, FSD architecture, Core Web Vitals | | testing | AI code failures, mutation testing, contract testing, flakiness |

AI providers

Set one or more API keys and primer will use the available provider:

export ANTHROPIC_API_KEY=sk-ant-...   # Claude (default preference)
export OPENAI_API_KEY=sk-...          # ChatGPT
export GEMINI_API_KEY=...             # Gemini

Override the model:

export ANTHROPIC_MODEL=claude-opus-4-6
export OPENAI_MODEL=gpt-4o-mini
export GEMINI_MODEL=gemini-2.0-pro

Configuration

primer.config.json in your working directory:

{
  "ai": {
    "maxTokens": 8192,
    "maxAgents": 4,
    "maxCommandsPerAgent": 3
  }
}

Or primer.config.mjs for dynamic config:

export default {
  ai: { maxTokens: 16384, maxAgents: 6 }
};

Full documentation

See the primer monorepo for full documentation, development setup, and release process.