skills

skills

Agent skill factory to create skills for feature and project development.

Each skill/feature lives underskills/<name>/SKILL.md, following the Agent Skills spec.

• skills/ - current skills and routed feature skills
• specs/ - generated specs
• loops/ - generated ralph loops

Install

Claude Code

/plugin marketplace add mattwoodco/skills
/plugin install skills@mattwoodco

Cursor

Install from this repository:

/add-plugin https://github.com/mattwoodco/skills

Cursor will read the manifest from .cursor-plugin/plugin.json.

Install locally while developing this plugin:

/add-plugin /Users/mw/Developer/mattwoodco/skills

For local installs, Cursor reads .cursor-plugin/plugin.json from that folder.

Local dev when /add-plugin doesn't load: Open this repo as your Cursor workspace (File → Open Folder → skills). The agent sees AGENTS.md, skills/, and .cursor-plugin/ directly, so you can edit manifests and skills and run validations here. To use these skills from another project, open that project in Cursor and @-mention the skill paths (e.g. @skills/add-feature/SKILL.md) or wait for Marketplace/install to work.

Marketplace submission guide: docs/cursor-marketplace-approval.md.

1. Entry skills
   • add-project: start a new project from an idea
   • add-feature: add one thing to an existing project
   • add-spec: write the spec only
2. Feature skills
   • building blocks like db, auth, ai-chat, payments, realtime
   • used by the entry skills in dependency order
3. Validation and operations
   • test-skills, improve-skill, eval-projects, ralph-ci

---

Skills (registered)

Skill	Purpose
add-spec	Write a project spec
add-feature	Feature development (routes to other skills)
add-project	Build a new project from idea -> spec -> shell

---

Features (routed)

Features are written in skill format and become available when routed to by the entry skills. They are not registered in the plugin's top-level skills list.

Project setup

Feature	Description
create-next	Initialize a new Next.js project with optimal defaults
add-shadcn	Design system using shadcn/ui
setup-vercel	Setup Vercel for a project
setup-lefthook	Git hooks with biome linting and type checking
env-config	Type-safe environment configuration with validation
env-from-1password	Copy environment variables from 1Password to .env.local

AI / LLM

Feature	Description
ai-core	Foundation AI layer - configures AI Gateway and exports getModel()
ai-sdk	AI SDK integration
ai-chat	Streaming chat UI with Postgres persistence
ai-tools	Tool calling framework with collapsible UI cards
ai-artifacts	Rich artifact panel - code, HTML, Mermaid, tables
ai-reasoning	Extended thinking display with collapsible accordion
ai-memory	Cross-session memory with context injection
ai-tasks	Task extraction and management from conversation
ai-mcp	Config-driven MCP server integration
ai-multi-agent	Multi-agent orchestration - N agents debate sequentially
ai-generative-ui	Tool results rendered as React components
ai-image-gen	AI image generation with Replicate and fal.ai
ai-video-gen	AI video generation - text-to-video, image-to-video
ai-voice-room	AI voice agent for LiveKit video rooms
ai-meeting-notes	Summaries, decisions, action items
ai-captions	Live closed captions via Deepgram + LiveKit
workflow	Durable AI agent infrastructure with Workflow DevKit

RAG