Overview
Agent Skills is a repository of engineering skills written by Addy Osmani that an AI coding agent loads as instructions. Each skill encodes a workflow, a quality gate or a set of practices — test-driven development, API and interface design, frontend UI engineering, code review — so the agent applies the same discipline on every task instead of improvising a different approach each session.
The pack is organised around the development lifecycle. Nine slash commands map onto the phases — `/spec` to define what to build, `/plan` to break it into small atomic tasks, `/build` to implement one slice at a time, `/test` to prove it works, `/constraints` to set the quality bar, `/review` before merge, `/webperf` to audit web performance, `/code-simplify` to cut complexity, and `/ship` to release. Skills also activate on their own based on what the agent is doing: designing an API triggers the API-design skill, building UI triggers the frontend skill.
It is agent-agnostic. The quickest route is the open `skills` CLI (`npx skills add addyosmani/agent-skills`), which the README says installs into 70+ agents, but the repository also documents native integrations for Claude Code plugins, Cursor, the Antigravity CLI, Gemini CLI, Codex, OpenCode, GitHub Copilot, Windsurf, Kiro and Command Code. The project is MIT-licensed.
What it does
- Nine lifecycle slash commands — /spec, /plan, /build, /test, /constraints, /review, /webperf, /code-simplify, /ship
- `/build auto` generates the plan and implements every task in one approved pass, still test-driven and committing each task individually, pausing on failures
- Skills auto-activate from context — API work pulls in api-and-interface-design, UI work pulls in frontend-ui-engineering
- Installs into 70+ agents through the open skills CLI, or natively as a Claude Code plugin, Codex plugin, Gemini CLI skill or Antigravity plugin
- Individual skills can be installed on their own, e.g. code-review-and-quality, interview-me, test-driven-development
- MIT-licensed and readable — every skill is a plain Markdown file you can fork and edit
Getting started
The fastest path works with any agent: the open skills CLI copies the skill files into whatever tool you already use.
Install every skill with the skills CLI
One command installs all 25 skills into the agent you are running. Add --list first if you want to browse before committing.
npx skills add addyosmani/agent-skills # install all 25 skills
npx skills add addyosmani/agent-skills --list # browse before installingOr take a single skill
Each skill installs independently. Note the README's caveat: a per-skill npx install copies only skills/<name>/, not the repo-level references/ directory, so shared checklists are unavailable unless you install the whole repo.
npx skills add addyosmani/agent-skills --skill code-review-and-quality
npx skills add addyosmani/agent-skills --skill interview-me
npx skills add addyosmani/agent-skills --skill test-driven-developmentNative install in Claude Code
Register the repository as a plugin marketplace, then install the plugin.
/plugin marketplace add addyosmani/agent-skills
/plugin install agent-skills@addy-agent-skillsNative install in Codex, Gemini CLI or Antigravity
Each CLI has its own plugin or skill installer; the repository ships the layout each one expects.
codex plugin marketplace add addyosmani/agent-skills
codex plugin add agent-skills@agent-skills
gemini skills install https://github.com/addyosmani/agent-skills.git --path skills
agy plugin install https://github.com/addyosmani/agent-skills.gitRun a phase
Once installed, drive the agent through the lifecycle by command. Each one activates the relevant skills automatically.
/spec # define what to build
/plan # break it into small atomic tasks
/build # implement one slice at a time
/test # prove it works
/review # improve code health before merge
/ship # releaseCommands and code are distilled from the project's own documentation — always check the official repo for the latest.
When to use it
- Give a coding agent a repeatable spec → plan → build → test → review → ship workflow instead of ad-hoc prompting
- Enforce a shared quality bar across a team whose members use different agents (Claude Code, Cursor, Codex, Copilot)
- Audit a site's web performance through the agent with /webperf rather than a separate tooling pass
- Borrow one practice — test-driven development or five-axis code review — without adopting the whole pack
How Addy's Agent Skills compares
Addy's Agent Skills alongside other open-source agent skills & plugins tools AI/TLDR tracks, ranked by GitHub stars.
| Tool | Stars | What it does |
|---|---|---|
| Superpowers | ★ 284k | A composable skills plugin that installs a spec-first, TDD, subagent-driven development methodology into Claude Code, Codex, Cursor, Gemini CLI and other coding harnesses. |
| Skills for Real Engineers | ★ 258k | Matt Pocock's everyday agent skills for coding agents, covering alignment grilling, planning, code review and research — small, composable and meant to be edited. |
| ECC | ★ 255k | An installable plugin that adds 68 agents, 286 skills, hooks, rules, memory and an agent-config security scanner to Claude Code, Codex and other coding harnesses. |
| Addy’s Agent Skills | ★ 93.1k | A set of engineering skills and slash commands that walk an AI coding agent through spec, plan, build, test, review and ship |
| Humanizer | ★ 45.8k | An agent skill that rewrites AI-sounding prose against a list of 25 named writing patterns, grouped into five sections and applied without changing the claims in the text. |
| Claude for Financial Services | ★ 34.8k | Anthropic reference agents, skills and MCP connectors for investment banking, equity research, private equity and wealth management, as Cowork plugins or Managed Agents. |
| Anthropic-Cybersecurity-Skills | ★ 32.5k | A community-built library of 800+ agentskills.io-format cybersecurity skills for AI agents, each mapped to MITRE ATT&CK, NIST CSF 2.0, ATLAS, D3FEND, NIST AI RMF and MITRE F3. |
| ARIS | ★ 15.9k | A Markdown-only skill pack that turns a coding agent into an autonomous research loop, with a second model acting as adversarial reviewer and a persistent research wiki. |