全部 skills

superpowers

obra/superpowers

Agentic skills framework and software development methodology that works (TDD, systematic debugging, planning, verification)

ponytail

DietrichGebert/ponytail

Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.

caveman

JuliusBrussee/caveman

Viral skill + proxy for coding agents that cuts 65% of tokens by talking like a caveman.

agent-skills

addyosmani/agent-skills

Production-grade engineering skills for AI coding agents.

claude-mem

thedotmack/claude-mem

Persistent context across sessions for every agent - captures everything your agent does during sessions.

Understand-Anything

Egonx-AI/Understand-Anything

Graphs that teach > graphs that impress. Turn any code into an interactive knowledge graph you can explore.

oh-my-openagent

code-yeongyu/oh-my-openagent

OmO: graph engineering skills for AI coding agents.

Code Review & Quality

addyosmani/agent-skills

Multi-axis code review with severity gates (Critical/High/Medium/Low) across correctness, readability, architecture, security, and performance. By Addy Osmani (Google Chrome). Nothing merges without a verdict.

Claude Code · Codex · Cursor · Copilot

Diagnosing Bugs

mattpocock/skills

A battle-tested diagnosis loop for hard bugs and performance regressions: reproduce → minimise → hypothesise → instrument → fix → regression-test. Includes a human-in-the-loop helper script. From a working engineer's daily setup.

Claude Code · Codex · Cursor

SwiftUI Pro

SwiftUI Pro

Teaches your agent to review and write modern SwiftUI: deprecated-API checks, data flow, navigation, accessibility and performance, by Paul Hudson of Hacking with Swift.

Claude Code · Codex · Cursor · Copilot +3

Readonly Question

Readonly Question

Ask anything about a codebase without changing a single file. A read-only /question mode for coding agents: evidence with file and line numbers, a proposed diff that is written out but never applied, and a verifiable 'No files were changed' footer. Understand the code first — change it only when you explicitly say so.

Claude Code · Codex · Cursor · Copilot +3

Systematic Debugging

obra/superpowers

Stop guessing at fixes. A 4-phase root-cause discipline (reproduce → isolate → hypothesize → verify) that forces your coding agent to find the real cause before touching code. The most-copied debugging skill in the Agent Skills ecosystem.

Claude Code · Codex · Cursor · Copilot +1

Incident Postmortem Writer

Incident Postmortem Writer

Write a blameless incident postmortem a manager can trust: a verified timeline, quantified impact, a root cause that survives five whys, and action items with one owner and one date each.

Claude Code · Codex · Cursor · Copilot +3

Tidy — AI Code Review

mblode/agent-skills

A diff/PR reviewer built specifically for agent-written code. Catches AI-slop patterns that line-by-line reviews miss, tiers findings as confirmed vs plausible, and can apply the fixes itself. The differentiator in your review stack.

Claude Code · Codex · Cursor

Test-Driven Development

obra/superpowers

RED-GREEN-REFACTOR with iron laws: write the test first, watch it fail, then write the minimal code to pass. Turns your agent into a disciplined TDD practitioner that never ships untested code.

Claude Code · Codex · Cursor · Copilot +1

Git Workflow & Versioning

addyosmani/agent-skills

Branch strategy, commit discipline, PR hygiene, and release versioning for agent-driven development. Keeps AI-generated changes manageable, reviewable, and reversible at high speed.

Claude Code · Codex · Cursor · Copilot +1