cmb-ai-kit

AI agent configuration kit -- rules and skills for Cursor and Claude Code, organized by domain.

npx cmb-ai-kit add       # interactive: pick env + rules, skills, agents
npx cmb-ai-kit list      # show all rules, skills, and agents

How it works

domains/                  <-- core rules/skills/agents (curated by maintainers)
  {domain}/
    rules/*.md            <-- concise guardrails (<50 lines)
    skills/*/SKILL.md     <-- detailed workflows (<500 lines)
    agents/*.md           <-- subagent definitions (Cursor / Claude Code)

workspaces/               <-- contributed rules/skills/agents (PR-based)
  {author}/
    rules/*.md
    skills/*/SKILL.md
    agents/*.md

npx cmb-ai-kit build cursor  -->  .cursor/rules/*.mdc + .cursor/skills/*/ + .cursor/agents/*.md
npx cmb-ai-kit build claude  -->  CLAUDE.md + .claude/agents/*.md

Commands

| Command | Description | |---------|-------------| | cmb-ai-kit add | Interactive installer: select environment, pick domains, install | | cmb-ai-kit build <cursor\|claude\|all> | Non-interactive: install everything for a target | | cmb-ai-kit list | Show all domains, rules, skills, and agents | | cmb-ai-kit clean | Remove generated outputs |

All commands accept --target <dir> to specify the project directory (defaults to .).

Domain index

| Domain | Rules | Skills | Agents | |--------|-------|--------|--------| | dev | coding-standards, testing-strategy, governance | bug-hunter, code-review | verifier, debugger | | arch | system-design | architect | -- | | security | baseline, secrets | threat-modeler | -- | | docs | standards | docs-expert, adr-writer | -- | | ai | eval-gates, stability-slo, cost-budget | evals-builder, mlops-systems, mcp-builder, workspace-evaluator | -- | | devops | ci-cd, deployment-safety | -- | -- | | data | modeling, migration-safety | data-schema-designer | -- | | frontend | standards | frontend-design | -- | | api | design | api-spec-builder | -- | | pm | jira-ticket | sprint-automation, subagent-playbooks | -- |

17 rules, 15+ skills, 2 agents across 10 domains (plus workspaces).

Source file format

Rules use YAML frontmatter:

---
description: What this rule enforces
alwaysApply: false
---

Skills use the same convention:

---
name: skill-name
description: What this skill does and when to trigger it
---

Subagents (agents) use markdown with YAML frontmatter compatible with Cursor and Claude Code:

---
name: agent-name
description: When to delegate to this subagent (used for automatic delegation)
model: fast | inherit | sonnet | haiku
---

The body is the system prompt for the subagent.

Adding core assets

1. Create domains/{domain}/rules/{topic}.md, domains/{domain}/skills/{name}/SKILL.md, or domains/{domain}/agents/{name}.md.
2. Run npx cmb-ai-kit build all (or node bin/cli.js build all locally).

Contributing workspaces

1. Create workspaces/{your-name}/rules/{topic}.md, workspaces/{your-name}/skills/{name}/SKILL.md, or workspaces/{your-name}/agents/{name}.md.
2. Open a PR. Workspace assets appear under a separate section in the interactive browser.

Local development

npm install
node bin/cli.js list
node bin/cli.js build all
node bin/cli.js add