GITHUB AGENT SKILLS / 技能收录
给你的 AI 助手
装上专业技能
我们从 GitHub 持续收录、审核并分类优质的 Agent Skills,按职业与使用场景整理,复制一行命令即可安装到 Claude Code、Cursor 等智能体。
按职业浏览 Agent Skills
为不同岗位挑选的高频技能
按使用场景浏览 Agent Skills
从你要完成的任务出发
按创作者浏览 Agent Skills
按 GitHub Stars 排序的优秀创作者
Microsoft
@microsoftopenclaw
@openclawAnthropic
@anthropicsOpenAI
@openaiThe Apache Software Foundation
@apacheGitHub
@githubAutoGPT
@Significant-GravitasGarry Tan
@garrytan全部 Skills 482556
local-files
Development-file storage rule for skills that write plans, reports, specs, handovers, or work ledgers. Names the directories Quiver writes and how each one is kept out of git on first write, unless the directory is already tracked or the project says to track it.
load-handover
Load the most recent handover note from the previous session into context.
hypothesis-debugging
Hypothesis-first debugging -- collects symptoms, generates and tests hypotheses with conditional agent dispatch, falls back to adaptive exploration, and proposes reviewed fixes.
handover
Summarize the current work state and prepare a handover note for the next session.
design
Extract a Figma design into a pixel-exact implementation plan -- reads the selected nodes through the figma-bridge MCP, maps Figma variables onto the project's existing theme tokens, and writes a self-contained plan to .claude/plans/ that /design-build executes without touching Figma again. --auto carries the same run through the build without a further prompt.
design-fix
Compare one already-implemented screen or component against its Figma source and fix the deviations you pick -- reads the node live through the figma-bridge MCP, or from the Node Specs of an existing /design plan, then diffs box, layout, typography, fill, stroke, radius, effects, and content against the code that renders it. Reports first, fixes only what you select, and never touches Figma.
design-build
Execute a design plan produced by /design -- implements each node against its embedded measurement spec and gates every task on the project's build or tests under a bounded retry budget. Runs with Figma disconnected; the plan carries every number it needs. --auto runs the whole loop without a prompt.
create-pr
Create a GitHub pull request from the current branch. Use when the user says 'create a pr', 'open pr', 'push and open a pr', 'create pull request', or wants to open a PR from their branch.
create-agents-md
Generate or rewrite an AGENTS.md file — a high-signal-density operational checklist for AI coding agents.
commit
Generate a Conventional Commits message, commit, and optionally push to remote.
code-navigation
CodeGraph-first code navigation with LSP and grep fallback. Provides agents with a reusable strategy for semantic code exploration -- codegraph tools when available, LSP goToDefinition/findReferences/documentSymbol otherwise, grep as last resort.
brainstorm
Explore ideas, compare approaches, and produce a validated spec before planning. Use when the user describes what to build, asks how to approach something, proposes a new feature or design change, or needs to explore options before implementation.
advise
Get a senior-style review of code, a planned change, a library choice, or a fix idea -- without writing a spec document. Multi-turn chat, grounded in your actual code. Use when you want a quick second opinion, not a plan.
casefile
Use when tracking security investigations, bug bounty findings, CTF leads, audit evidence, exploit chains, dead ends, or reports in the Casefile ledger through Codex.
chrome-performance
Use Chrome DevTools MCP for frontend performance work such as slow page loads, LCP, CLS, INP, Lighthouse regressions, scripting hotspots, or layout thrash. Trigger when the user asks for page speed analysis, Core Web Vitals debugging, trace-based investigation, or before-and-after performance verification.
chrome-devtools
Use Chrome DevTools MCP to inspect live pages, reproduce browser issues, automate user flows, inspect console and network activity, and verify frontend fixes in a real Chrome session. Trigger when the user mentions browser bugs, client-side errors, broken forms, flaky UI behavior, or asks to interact with a site directly.
unattended
Starts the Clean Room startup wizard in bounded unattended controller mode for authorized source-to-implementation work with finite loop limits and safety stops.
start-over
Archives or quarantines current Clean Room artifacts and restarts from the scope gate with a fresh task id.