AI Skills Hub

A community-driven collection of reusable AI skills — for discovering, exploring, and reusing specialized agent capabilities.

npx github:codebygarv/Ai-skills add grill-me

That's it — one command drops a working skill into .claude/skills/grill-me/ in your project.

See it in action

GrillMe — challenges an idea instead of agreeing with it.

You: "Grill me on this: our new feature flag system stores flags as rows in the primary Postgres database, read on every request."

GrillMe: Every request now takes a read-path dependency on your primary database for something that's almost always the same answer for minutes at a time — you've turned a config lookup into a scaling bottleneck for your most critical path.

1. Assumption: flag reads are cheap. At 10x traffic, this competes for connections with actual business queries — a flag check can degrade checkout.
2. No caching layer. A single flag toggle now means every pod issues fresh reads with zero staleness tolerance, when 30 seconds of staleness would've been invisible.
3. Alternative not addressed: a push-based cache or a dedicated flag service solves this entirely — why wasn't that considered?

Questions to answer before shipping this: What's your target read latency at peak, and does this design hit it with headroom?

Full transcript: skills/reasoning/grill-me/examples/example.md. 49 more skills below.

What is this?

AI Skills Hub is a centralized catalogue of reusable AI skills designed to extend the capabilities of AI coding agents, assistants, and developer workflows.

Instead of every developer re-writing the same instructions, prompts, and agent behaviors from scratch, you can browse this catalogue, find a skill that matches what you need, read what it does, and drop it into your own AI workflow.

Each skill is a self-contained package of instructions that teaches an AI agent how to perform one specific type of task consistently — for example:

• GrillMe → aggressively reviews, questions, and challenges an idea, implementation, architecture, or plan.
• UI Auditor → analyzes a UI implementation for design-system, responsiveness, accessibility, and consistency issues.
• PR Reviewer → reviews a pull request like a senior engineer and flags bugs, architectural problems, and maintainability issues.
• UI Text Formatter → detects poorly formatted UI text and improves capitalization, spacing, hierarchy, and readability.

Core Vision

Think of it like a package registry, but for AI agent skills:

AI Skills Hub
      ↓
Search / Explore Skills
      ↓
Find "PR Reviewer"
      ↓
Read Documentation
      ↓
Install / Copy Skill
      ↓
Use it with your AI Agent

The long-term goal is to make AI skills discoverable, reusable, shareable, version-controlled, community-driven, easy to install, easy to customize, and easy to contribute to.

What is an AI Skill?

An AI Skill is a structured set of instructions that gives an AI agent a specialized capability or behavior. A skill defines:

• What the AI should do
• When the AI should use the skill
• What information it should analyze
• What rules it should follow
• What output it should produce
• What it should avoid
• Examples of expected behavior

See docs/SKILL_SPEC.md for the full folder structure and quality standards every skill in this repo follows.

The Catalogue

50 skills across 9 categories.

🧠 Reasoning (skills/reasoning/)

💻 Development (skills/development/)

🎨 UI / UX (skills/ui-ux/)

📝 Documentation & Communication (skills/documentation/)

🧪 Quality & Testing (skills/testing/)

🚀 Project & Architecture (skills/architecture/)

🔐 Security (skills/security/)

🚀 Productivity (skills/productivity/)

🔧 Utilities (skills/utilities/)

Standard Skill Structure

Every skill follows the same layout:

skills/<category>/<skill-name>/
├── SKILL.md          # Instructions the AI agent actually reads (YAML frontmatter + body)
├── README.md          # Human-facing docs: what it does, when to use it, examples
└── examples/
    └── example.md     # A worked example of the skill in action

Full spec: docs/SKILL_SPEC.md. Contribution steps: CONTRIBUTING.md.

Using a Skill

Option 1 — CLI (recommended): install directly from this repo with npx, no clone or npm publish required.

# install one skill
npx github:codebygarv/Ai-skills add grill-me

# install several at once
npx github:codebygarv/Ai-skills add pr-reviewer security-auditor

# see everything available
npx github:codebygarv/Ai-skills list

# install somewhere other than the default ./.claude/skills
npx github:codebygarv/Ai-skills add grill-me --target .ai/skills

By default this drops each skill into .claude/skills/<skill-name>/ in your current directory — Claude Code picks up skills from there automatically. Point --target elsewhere for other agents/conventions.

Option 2 — manual: browse the catalogue below, open a skill's README.md, and copy the skill's folder into your agent's skills directory yourself.

Cloned this repo locally instead? Run the same commands with node bin/ai-skills.js in place of npx github:codebygarv/Ai-skills.

Composing Skills

Skills are designed to be composable — stack them for deeper workflows:

GrillMe + Architecture Reviewer + Security Auditor = Deep Technical Review

Roadmap

• Phase 1 — Foundation: repo structure, skill spec, 50 skills across 9 categories, root docs, contribution guidelines, CI test suite.
• Phase 2 — Discovery: skill catalogue site, categories/tags/metadata, search, popularity indicators.
• [~] Phase 3 — Installation: add/list CLI usable via npx github:codebygarv/Ai-skills ✅ — still to do: npm registry publish, update/remove commands, versioning.
• Phase 4 — Community: community submissions, reviews, ratings, discussions, featured/trending skills.

Development

No dependencies — the CLI and tests use only the Node standard library (Node 18+).

npm test     # run the test suite (Node's built-in runner)
npm run lint # validate every skill's structure and frontmatter
npm run check # both

Tests cover the CLI end-to-end (install, multi-install, --target, error paths, exit codes) and catalogue integrity — every skill matches docs/SKILL_SPEC.md, skill names are unique, and the README's catalogue links and skill counts stay in sync with what's actually on disk. Both run in CI on every push and PR.

Contributing

New skills, fixes, and improvements are welcome — see CONTRIBUTING.md for the process and quality bar.

Long-Term Vision

Instead of every developer maintaining their own private collection of prompts and agent instructions, they discover specialized, vetted skills from one shared ecosystem — the place you go whenever you think "I need my AI agent to be really good at this specific task."