@loomcraft/cli

AI workflow preset manager — 7 universal agents, skills.sh integration, 10 presets.

Loomcraft scaffolds production-ready presets into your project. Each preset bundles 7 stack-agnostic agents, skills from the skills.sh ecosystem, and a workflow pipeline with TDD hybrid patterns.

Quick start

# Interactive — pick a preset or agents
npx @loomcraft/cli add

# One-liner with a preset
npx @loomcraft/cli add loomcrafthq/presets/saas
npx @loomcraft/cli add loomcrafthq/presets/api

Commands

loomcraft add [ref]

Add a preset or agent from a GitHub repo.

loomcraft add                              # interactive wizard
loomcraft add loomcrafthq/presets/saas     # add the SaaS preset
loomcraft add loomcrafthq/agents/backend   # add a single agent

Flag	Description
--agent	Force treating ref as an agent
--preset	Force treating ref as a preset
--claude	Use Claude Code target (.claude/ + CLAUDE.md)
--cursor	Use Cursor target (.cursor/ + .cursorrules)
--target <name>	Output target (claude-code, cursor, custom)
--target-dir <dir>	Custom target directory
--context-file <file>	Custom context file name
--overwrite	Overwrite existing context file instead of merging

Context file merge: When CLAUDE.md already exists, add preserves your custom content and only updates the Loomcraft-managed sections. Use --overwrite to replace the entire file.

loomcraft find [query]

Search community presets and agents.

loomcraft find                # list all
loomcraft find tailwind       # search by keyword
loomcraft find --type agent   # agents only

Flag	Description
--type <type>	Filter: agent or preset
--sort <sort>	Sort: popular, recent (default: popular)

loomcraft sync

Re-install skills and update the context file.

loomcraft sync

Flag	Description
--target <name>	Output target
--target-dir <dir>	Custom target directory
--context-file <file>	Custom context file name

loomcraft clean

Remove all files created by Loomcraft.

loomcraft clean
loomcraft clean --dry-run    # preview without deleting

Flag	Description
--dry-run	Preview without deleting
--target <name>	Output target
--target-dir <dir>	Custom target directory
--context-file <file>	Custom context file name

What's generated

Running loomcraft add loomcrafthq/presets/saas creates:

.claude/
  agents/
    database/AGENT.md    # schema design, migrations, queries
    backend/AGENT.md     # APIs, auth, business logic
    frontend/AGENT.md    # components, state, accessibility
    ux-ui/AGENT.md       # design system, usability, responsive
    tester/AGENT.md      # unit, integration, e2e, TDD
    review-qa/AGENT.md   # code review, quality assurance
    security/AGENT.md    # OWASP Top 10, audit, hardening
  skills/
    brainstorming/SKILL.md
    ticket-craft/SKILL.md
    ...16 more            # installed from skills.sh
CLAUDE.md                 # project context (auto-read by Claude Code)
skills.json               # skills.sh references

Claude Code discovers agents in .claude/agents/ automatically. The skills.json file references external skills from the skills.sh ecosystem.

Agents (7)

All agents are stack-agnostic. They read your project's CLAUDE.md to adapt to whatever framework, language, or tooling you use.

Agent	Role
database	Schema design, migrations, query optimization, data integrity
backend	API design, business logic, auth, clean architecture
frontend	UI components, state management, accessibility (WCAG 2.1 AA), performance
ux-ui	Design system, usability heuristics, responsive design, animation
tester	Unit/integration/e2e tests, TDD, Arrange-Act-Assert
review-qa	Code review, quality audit, security and performance checks
security	OWASP Top 10, vulnerability audit, secure coding patterns

Skills (skills.sh)

Skills are external packages from the skills.sh ecosystem. Each preset references the skills it needs in skills.json.

Core skills (included in all presets)

Skill	Source
Brainstorming	obra/superpowers/brainstorming
Writing Plans	obra/superpowers/writing-plans
Executing Plans	obra/superpowers/executing-plans
Systematic Debugging	obra/superpowers/systematic-debugging
Code Review	obra/superpowers/requesting-code-review
Receiving Review	obra/superpowers/receiving-code-review
Conventional Commits	github/awesome-copilot/conventional-commit
Ticket Craft	loomcrafthq/skills/ticket-craft
Testing Patterns	loomcrafthq/skills/testing-patterns

Ecosystem skills (per preset)

Skill	Source
TDD	obra/superpowers/test-driven-development
Frontend Design	anthropics/skills/frontend-design
Web App Testing	anthropics/skills/webapp-testing
Web Design	vercel-labs/web-design-guidelines
React Best Practices	vercel-labs/react-best-practices
Next.js Best Practices	vercel-labs/next-best-practices
React Native	vercel-labs/react-native-skills
Better Auth	better-auth/skills
Supabase	supabase/agent-skills

Workflow Pipeline

Each preset defines a workflow with 4 phases:

1. PREPARATION → 2. PIPELINE → 3. VERIFICATION → 4. FINALIZATION

Example: TDD Hybrid Pipeline

workflow:
  preparation:
    source: linear          # ticket source (linear, github-issues, manual)
  pipeline:
    - agent: database
    - agent: tester
      mode: tdd             # write tests first (red)
    - agent: backend
    - agent: ux-ui
    - agent: frontend
    - agent: tester
      mode: test-after      # write tests after implementation
  verification:
    - review-qa
    - security
  finalization:
    commits: conventional
    branch: feat/{{ticket}}

Presets (10)

Preset	Use case
saas	Full-stack SaaS with auth, billing, TDD pipeline
api	Backend API service
landing	Marketing landing page
mobile	React Native / Expo mobile app
ecommerce	Online store with Stripe
cli	Node.js command-line tool
extension	Chrome browser extension
ai-app	AI-powered app (chatbot, RAG)
realtime	Real-time app (chat, collaboration)
blog	Blog / CMS with MDX

Every preset includes the 7 universal agents and core skills.sh references.

License

MIT