Source-verified agent skills

AI agent skills forbetter coding workflows.

Find source-verified coding skills for frontend design, React and Next.js performance, browser testing, security review, MCP development, and repeatable engineering work.

8

Launch collection

3

Official publishers

100%

Source and terms visible

Curated modules

Add a repeatable capability

Each profile explains what the skill changes, when it earns its context, how to install it, and what to inspect before trusting it.

8 skills

01
Design Recommended
frontend-design

Frontend Design

Give an agent a stronger visual point of view before it builds or restyles a web interface.

UIArt directionTypography

Anthropic

Low setup · Claude Code

View skill
02
Design Recommended
web-design-guidelines

Web Design Guidelines

Audit UI code against Vercel-maintained interface guidance for accessibility, UX, content, and performance.

AccessibilityUX reviewUI quality

Vercel Labs

Low setup · Codex

View skill
03
Engineering Recommended
react-best-practices

React Best Practices

Apply prioritized React and Next.js performance guidance from Vercel Engineering while writing or reviewing code.

ReactNext.jsPerformance

Vercel Labs

Low setup · Codex

View skill
04
Engineering
mcp-builder

MCP Builder

Guide the design and implementation of Model Context Protocol servers with useful, well-scoped tools.

MCPToolsIntegrations

Anthropic

Some setup · Claude Code

View skill
05
Quality Recommended
webapp-testing

Web App Testing

Test local web applications through real browser flows, screenshots, console output, and interaction checks.

Browser testingPlaywrightDebugging

Anthropic

Some setup · Claude Code

View skill
06
Quality
security-review

Security Review

Trace trust boundaries and risky data flows through a codebase using GitHub’s reusable review workflow.

SecurityCode reviewData flow

GitHub

Some setup · GitHub Copilot

View skill
07
Workflow Recommended
acquire-codebase-knowledge

Acquire Codebase Knowledge

Map an unfamiliar repository and turn scattered implementation details into a useful onboarding view.

OnboardingArchitectureDocumentation

GitHub

Low setup · GitHub Copilot

View skill
08
Workflow
skill-creator

Skill Creator

Create or improve reusable Agent Skills with clear triggers, scoped instructions, and supporting resources.

Skill authoringEvaluationAutomation

Anthropic

Low setup · Claude Code

View skill

Build a stronger agent workflow

Pair a clear build prompt with a focused skill.

Use a prompt to define the product outcome, then add the skill that improves how your agent designs, implements, tests, or reviews the work.

How to choose an agent skill

Install for a recurring job, not for the size of the catalog

Every installed skill competes for attention and can influence agent behavior. Start with one workflow you repeat, verify that the trigger is specific, and keep only skills that measurably improve the result.

  1. 01

    Name the repeated task

    Choose a concrete job such as UI review, browser testing, codebase onboarding, or MCP server design.

  2. 02

    Inspect the source

    Read the trigger description, instructions, scripts, references, dependencies, and license terms.

  3. 03

    Check agent compatibility

    Confirm the expected folder, install command, tools, runtime, and permission model for your coding agent.

  4. 04

    Run a controlled evaluation

    Compare a realistic task with and without the skill, then keep it only if the improvement is clear.

Prompt vs skill

A prompt starts a task. A skill standardizes how it gets done.

Use a prompt for one outcome. Reach for a skill when the workflow repeats, needs bundled references or scripts, or should activate consistently across a team.

Browse build prompts
01

Instructions

A focused SKILL.md explains triggers, steps, constraints, and expected output.

02

Resources

Scripts, references, templates, and assets can be loaded only when needed.

03

Repeatability

The same review or build method can travel between projects and compatible agents.

Before installation

Read the skill like executable policy.

A skill can influence commands, files, network access, and decisions. Popularity is useful discovery evidence, but it is not a security boundary.

  1. 01

    Inspect every file

    Read SKILL.md and bundled scripts, not just the repository description.

  2. 02

    Check permissions

    Understand which commands, files, credentials, and network destinations it can touch.

  3. 03

    Pin the source

    Prefer a reviewed commit or release when reproducibility matters.

  4. 04

    Test in isolation

    Use a disposable branch or workspace before trusting it with a production repository.

Frequently asked questions

AI agent skills, explained

Understand what coding skills contain, which agents can use them, how they differ from prompts, and what to review before installation.

What is an AI agent skill?

An AI agent skill is a folder of reusable instructions centered on a SKILL.md file. It teaches a compatible coding agent when and how to perform a specific workflow and may include scripts, references, templates, or other supporting assets.

How are agent skills different from prompts?

A prompt asks for one outcome in the current conversation. A skill defines a repeatable method that can be discovered or invoked when relevant. Use prompts to describe what you want to build and skills to standardize how design, testing, review, or another recurring task should be performed.

Which coding agents support skills?

Support and installation locations vary. Claude Code, GitHub Copilot, Codex, Cursor, and other tools increasingly support the Agent Skills format or compatible skill folders. Check each profile and the current documentation for your agent before installing.

Are community agent skills safe to install?

Do not assume so. A skill can influence commands, file changes, network access, and use of credentials. Read SKILL.md and every bundled script, inspect requested permissions, prefer a reviewed commit, and test in a disposable branch or workspace.

Do agent skills add new tools or permissions?

A skill primarily adds instructions and resources; it does not magically grant capabilities the agent does not have. However, those instructions may ask an existing shell, browser, or integration tool to perform actions, so the agent’s available permissions still matter.

Can I create a private skill for my team?

Yes. Team-specific skills are useful for release checks, code review standards, migrations, documentation, and other repeatable procedures. Keep secrets out of the skill, document prerequisites, test realistic tasks, and store it where only the intended users or repositories can access it.

Start with one useful capability

Choose the workflow bottleneck you want to remove.

Open a verified skill profile, inspect the original source, copy the installation command, and test it in a safe project before rolling it out to your team.