AI Agent Skills
A library of 30-plus instruction skills for coding agents: planning, building, testing, hallucination guards and shipping
- 217 ★ on the repository
- Last commit
Minimal access
Why this level
- SKILL.md files hold instructions only and run nothing, but they change the agent's behavior in your project
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/developersglobal-ai-agent-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- SKILL.md files hold instructions only and run nothing, but they change the agent's behavior in your project
This is third-party code. Review the repository files before installing.
What it does
AI Agent Skills is a collection of SKILL.md files that hold working procedures with steps, checkpoints, verification gates and anti-pattern guards. The chain runs from idea to release: think, plan, build, test, review, harden, ship. The repository is inspired by Addy Osmani's and forrestchang's sets. Skills are grouped by stage. For thinking: think-before-coding, goal-driven-execution, idea-to-spec. For planning: task-decomposition, context-loading, multi-agent-orchestration. For building: incremental-coding, simplicity-first, surgical-changes, api-design and more. Then come testing, review, security (security-hardening, prompt-injection-defense, dependency-auditor, threat-modeler), deployment and observability, plus skills for memory and RAG and for everyday work such as meeting notes and email. Three agents live in agents (code-reviewer, security-auditor, test-engineer), accessibility and performance checklists sit in references, and a GitHub Pages browser searches and copies skills.
Who it is for. Developers who use Claude Code, Cursor, Gemini CLI or Copilot and want a disciplined process from clarifying the goal to testing and release.
Good fit when
- An agent needs rules for small verifiable steps and no large rewrites
- A task must become goals with success criteria and a testable spec
- Security hardening and prompt-injection defense run before release
- One set of instructions for Claude, Cursor and Gemini CLI is copied into rules files
Not a fit when
- You need an installer with versions and updates: skills are copied by hand into CLAUDE.md, .cursor/rules or GEMINI.md
- You need a narrow technology-specific set: the skills are general
- You need ready automation: the files give the agent instructions and do not run code
Example request
Read skills/think-before-coding/SKILL.md and follow it before changing the payments moduleLimitations
Skills are connected by hand: the SKILL.md contents are copied into CLAUDE.md, .cursor/rules or GEMINI.md or into a system prompt, with no automatic install or versioning. These are structured prompts and results depend on the model. Some files (skills/developer, skills/productivity) are standalone markdown with no SKILL.md folder. The comparison table in the README was written by the repository author, so read it critically.
How to disable. Remove the added fragments from CLAUDE.md, GEMINI.md and .cursor/rules or delete the copied SKILL.md files.
FAQ
Which agents are supported?
Any that reads instructions from a file: Claude Code (CLAUDE.md), Cursor (.cursor/rules), Gemini CLI (GEMINI.md), Copilot. The repository has AGENTS.md, CLAUDE.md and GEMINI.md for the project itself.
How do I add a skill to Claude Code?
Append it to the project's CLAUDE.md, for example cat skills/think-before-coding/SKILL.md >> CLAUDE.md. For Cursor copy it into .cursor/rules/ with the .mdc extension, and for Gemini CLI append to GEMINI.md.
What topics do the skills cover?
Planning and task breakdown, incremental building, APIs, debugging, refactoring, testing, review, security, CI/CD, deployment, observability, memory and RAG, hallucination and prompt-injection defense, plus meeting notes, email and learning plans.
What is in the agents folder?
Three agent definitions: code-reviewer, security-auditor and test-engineer. The references folder holds accessibility and performance checklists.
Is there a convenient browser?
Yes, an interactive skills browser on GitHub Pages (developersglobal.github.io/ai-agent-skills) with search and copy. The catalog is built from skills/index.json.
Discussion
No questions or reviews yet.
Related
SkillsThe author's personal set: shared agent rules and dozens of skills for GitHub, the browser, macOS, messaging and releases, plus a sync script
Over twenty skills for coding agents: MCP and hooks, plan mode, a bug-fix protocol, prompting and a bash safety hook
A set of independent skills against drift in large codebases: hard-cut, root-cause tracing, lean testing and CI right-sizing.
A skills library for Claude Code, Cursor and Codex: team code review, TDD, stacked PRs and repository readiness scoring
14 skills that find the gaps in a task before the code: interviews, plans, assumption tests and proof that a bug fix works
A skill that analyzes a repository and generates files for AI agents: AGENTS.md, pointers, reviewers, CI and templates
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| AI Agent Skills (this card) | no | 217 | |
| Agent Skills by Addy Osmani | no | 116,477 | Addy Osmani's set of engineering processes, while this one covers more: AI safety, RAG and everyday tasks |
| Tech Leads Club Agent Skills | no | 7,161 | a skill selection from a tech leads club, while this repository is split by stage from thinking to shipping |
README in short
The README explains the idea (instructions with working procedures), compares the set with two examples, gives a quick start with three ways of copying, lists skills by stage from thinking to shipping and shows the idea-to-release chain. It links to an interactive skills browser.
README badge
Original SKILL.md
--- name: think-before-coding description: Forces explicit reasoning before writing any code. Surfaces assumptions, manages confusion, and prevents hallucination by demanding clarity upfront. category: think applies-to: [claude, gemini, cursor, copilot, any] version: 1.0.0 --- ## Overview AI agents default to writing code immediately — even when the request is ambiguous, the context is incomplete, or multiple valid interpretations exist. This skill forces a deliberate thinking phase before any implementation begins. The result: fewer rewrites, fewer wrong-direction implementations, and fewer "I assumed X" surprises. ## When to Use Activate this skill **before writing any code** when: - The request is ambiguous or underspecified - Multiple valid implementations exist - You are unsure about an existing system's constraints - The task touches security, data integrity, or public APIs - You feel the urge to "just start coding" to make progress ## Process ### Phase 1: Understand the Request 1. **Read the full request** — Do not skim. Re-read once. 2. **Identify ambiguities** — List every decision you would have to make silently: - What counts as "success"? - What are the inputs and expected outputs? - What should NOT change? - Are there existing patterns to follow? 3. **Surface assumptions** — Write them out: *"I am assuming X because Y."* 4. **Check for contradictions** — Does the request contradict itself or existing code? **Verify:** Can you state the goal in one clear sentence? If not, ask before proceeding. ### Phase 2: Clarify Before Acting 5. If **genuinely ambiguous**, ask a focused clarifying question — not a list of 10 questions. Pick the one blocker. 6. If multiple valid approaches exist, **present 2–3 options with tradeoffs**, then ask which to proceed with. 7. If something seems wrong with the request, **say so directly** — don't silently work around it. **Verify:** You have a clear, agreed-upon interpretation of the task.