AI Skills Registry

Disclaimer: This project is experimental and a proof-of-concept. The CLI, registry format, and APIs are subject to breaking changes. Some skill definitions are placeholders and not fully fleshed out — they serve as examples of the intended structure rather than production-ready prompts.

A centralized registry and CLI for managing AI coding assistant skills. Keep prompts, context definitions, and strategy documents organized across agents, languages, themes, and services — installable in one command.

Supported agents: Google Antigravity · Claude Code · Cursor · Visual Studio Code (Copilot)

---

Features

• Multi-agent — skills for Antigravity, Claude Code, Cursor, and VSCode Copilot
• Rich metadata — each skill has a description, theme, tags, version, and author
• Themes — cross-cutting concerns: quality, architecture, security, dx
• Tags — find skills by framework or service (e.g. jest, owasp, pandas)
• Native install — install directly to each agent's config location (.claude/commands/, .cursorrules, etc.)
• Lockfile — pin team skills in .ai-skills.json and restore with one command
• Dynamic fetching — no local clone needed; fetches from GitHub at runtime

---

Quickstart

npm install
npm run build
npm link        # makes `ai-skills` available globally

---

Commands

list — browse available skills

ai-skills list

# Filter by any dimension
ai-skills list --lang typescript
ai-skills list --category security
ai-skills list --concept testing
ai-skills list --tag owasp

search — find skills by keyword

Searches across id, description, tags, concept, and agent name.

ai-skills search testing
ai-skills search "code review"
ai-skills search aws

info — inspect a skill before installing

ai-skills info testing --lang typescript
ai-skills info code-review --lang general

install — iInstall a specific skill into your project's ai-skills directory:

ai-skills install testing --lang typescript
ai-skills install security

To install directly into an agent's specific native config location, provide the --agent argument:

# Appends to .cursorrules
ai-skills install testing --agent cursor

# Creates .claude/commands/code-review.md
ai-skills install code-review --agent claude-code

# Appends to .agent/skills/git-workflow/
ai-skills install git-workflow --agent antigravity

install-all — restore from lockfile

# Create a lockfile
ai-skills init

.ai-skills.json example:

{
  "skills": [
    { "id": "testing/typescript" },
    { "id": "security/general" },
    { "id": "git-workflow/general", "agent": "claude-code" }
  ]
}

Then run install-all to download and apply them all at once. Skills with an agent specified will be installed natively for that agent. Skills without an agent will be installed in the default .ai-skills structure.

ai-skills install-all

---

Agent Native Install Locations

Agent	Strategy	Location
claude-code	Slash command	.claude/commands/<concept>.md
cursor	Append	.cursorrules
vsc	Append	.github/copilot-instructions.md
antigravity	Skill folder	.agent/skills/<concept>/

---

Skill Schema

Each skill in skills.json:

{
  "id": "testing/typescript",
  "concept": "testing",
  "language": "typescript",
  "category": "engineering",
  "tags": ["vitest", "jest", "tdd", "mocking", "coverage"],
  "description": "TypeScript testing with Vitest/Jest: file structure, mocking strategy, async patterns, and coverage targets",
  "version": "1.0.0",
  "author": "tpa",
  "files": ["SKILL.md"]
}

Themes: quality · architecture · security · dx

Skill files use YAML frontmatter matching the manifest entry, followed by the actual instructions in Markdown.

---

Available Skills

ID	Category	Tags
testing/typescript	engineering	jest, vitest, tdd
data-science/python	data	pandas, numpy, sklearn
code-review/general	engineering	review, owasp, readability
code-review/python	engineering	python, fastapi, pydantic
architecture/python	architecture	pep8, clean-architecture
testing/general	engineering	unit-testing, boundaries
refactoring/typescript	engineering	solid, dry, type-safety
git-workflow/general	engineering	git, conventional-commits, pr
security/general	security	owasp, injection, auth
package-management/python	engineering	uv, pip, venv
accessibility/general	auditing	accessibility, testing, ui
design/frontend	frontend	design, styling, ux
ci-cd/github-actions	devops	ci-cd, automation
project-management/jira	project-management	jira, agile
git-workflow/pr-generator	engineering	pr, automation
documentation/project	documentation	architecture, overview
documentation/readme	documentation	generative, docs
infrastructure/terraform	devops	iac, aws, azure, gcp

---

Adding New Skills

1. Create the directory: skills/<concept>/<language>/
2. Add a SKILL.md file with YAML frontmatter following the agentskills.io standard.
3. Add any necessary supplemental files, such as scripts/ or references/.
4. Add an entry to skills.json
5. Commit and push to main

The CLI fetches live from GitHub — no build or release step needed.

---

Configuration

Environment Variable	Default	Description
AI_SKILLS_REGISTRY_URL	GitHub skills.json	Custom manifest location
AI_SKILLS_REPO_URL	GitHub raw root	Custom repo base URL
AI_SKILLS_AGENTS_URL	GitHub agents.json	Custom agents config location