A collection of skills for AI coding agents. Skills are packaged instructions that extend agent capabilities with domain-specific knowledge and best practices.
Skills follow the Agent Skills format and are listed on the Agent Skills Directory.
Agent-agnostic branch-finishing workflow. Use refine when you want an agent to review the current branch, fix gaps, verify the branch, and ship it through commit, push, and PR update.
Use when:
- A branch is in progress and needs to be brought to review-ready quality
- You want review, lint, typecheck, tests, docs, and PR handling done in one pass
- You want the agent to fix issues, not only report them
- You want the branch committed, pushed, and synced with a PR
Can use these companion skills when available:
| Skill | Source | Install |
|---|---|---|
next-best-practices |
vercel-labs/next-skills | npx skills add vercel-labs/next-skills --skill next-best-practices |
next-cache-components |
vercel-labs/next-skills | npx skills add vercel-labs/next-skills --skill next-cache-components |
vercel-react-best-practices |
vercel-labs/agent-skills | npx skills add vercel-labs/agent-skills --skill react-best-practices |
vercel-composition-patterns |
vercel-labs/agent-skills | npx skills add vercel-labs/agent-skills --skill composition-patterns |
vercel-react-native-skills |
vercel-labs/agent-skills | npx skills add vercel-labs/agent-skills --skill react-native-guidelines |
web-design-guidelines |
vercel-labs/agent-skills | npx skills add vercel-labs/agent-skills --skill web-design-guidelines |
See skills/refine/ for the full branch-finishing workflow and usage examples.
Field-tested engineering principles. Contains 40+ rules across 12 categories, distilled from real project experience and hundreds of debugging sessions.
Use when:
- Writing any new code
- Making changes to existing code
- Debugging issues
- Reviewing code quality
- Before committing changes
- Working with AI coding assistants
Categories covered:
- Verification & Testing (Critical)
- Scope Discipline (Critical)
- Root Cause Analysis (High)
- Regression Prevention (High)
- Clarification Over Assumption (High)
- Staying Current (High)
- Simplicity & Performance (Medium)
- Quality Over Quantity (Medium)
- Workflow & Automation (Medium)
- Backward Compatibility (Medium)
- Security Awareness (Medium)
- AI Collaboration Principles (Medium)
Core philosophy: Never assume. Always verify.
Concrete wrong/correct examples of common AI coding agent mistakes. Derived from analysis of 500+ real chat sessions across 21 projects. Contains 25+ rules across 8 categories.
Use when:
- Writing code or modifying files
- Debugging issues
- Generating output
- Reviewing AI-generated code
Categories covered:
- File & Project Awareness (Critical)
- Multi-Target Instructions (High)
- Deprecated APIs (High)
- Output Formatting (Medium)
- Debugging Anti-Patterns (Medium)
- Wrong Optimizations (Medium)
- Content Integrity (Medium)
- Security Scope (Medium)
Complements agent-best-practices (principles) with specific, actionable anti-patterns.
Install all skills from this repo:
npx skills add sceiler/skills| Repository | Description |
|---|---|
| vercel-labs/agent-skills | Vercel's official skills — React best practices, web design guidelines, composition patterns, React Native, deploy |
| vercel-labs/next-skills | Next.js skills — best practices, cache components, upgrade guides |
Browse more skills at skills.sh.
Each skill contains:
SKILL.md— Instructions for the agent (required)metadata.json— Skill metadata (required)README.md— Human-readable documentation (optional)
MIT