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
@microsoft
openclaw
@openclaw
OpenAI
@openai
The Apache Software Foundation
@apache
GitHub
@github
Anthropic
@anthropics
AutoGPT
@Significant-Gravitas
Garry Tan
@garrytan全部 Skills 196921
dev
Development workflows for the playwright-cli repository. Use when the user asks about rolling dependencies, releasing, or other repo maintenance tasks.
dispatching-parallel-agents
Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies
digitalocean-mcp-setup
Use when installing, validating, or troubleshooting the archived DigitalOcean MCP server for Codex, Claude, or Gemini, especially when App Platform tools or env-only token wiring need to be set up safely.
digitalocean-app-platform-ops
Use when operating DigitalOcean App Platform through the archived DigitalOcean MCP server, especially for listing apps, deployments, logs, restarts, deletes, or spec validation.
diff-blind-test-synth
Use when you want an INDEPENDENT external check that a patch/PR actually satisfies the spec - synthesize executable tests from ONLY the spec + public API (never the diff), then run them against the pre-patch tree (expect FAIL) and the patched tree (expect PASS). A divergence flags a spec-miss or over/under-fitted patch. Distinct from fail-first/red-proof. SHADOW mode: never blocks.
devtu
ToolUniverse (mims-harvard/ToolUniverse) development toolkit - one skill, eight sub-command modes. Use when working on ToolUniverse tools: create a new scientific/API tool, fix a failing tool, optimize tool JSON descriptions, optimize research skills, auto-discover life-science APIs and generate integration PRs, run the debug-test loop, or push and fix CI. Each mode's playbook is under modes/.
develop-web-game
Use when Codex is building or iterating on a web game (HTML/JS) and needs a reliable development + testing loop: implement small changes, run a Playwright-based test script with short input bursts and intentional pauses, inspect screenshots/text, and review console errors with render_game_to_text.
design-tokens-pipeline
Use for the W3C DTCG design-token round-trip the TCR design-review loop relies on at G5/G6 - extract tokens out of code (CSS custom properties + Tailwind theme dumps) into a DTCG JSON tree, validate against the DTCG v1.0 rules that break the Tokens-Studio to Style-Dictionary round-trip, and diff two token trees to emit the forward change set AND the inverse revert patch.
design-system
Use this skill to generate or audit design systems, check visual consistency, and review PRs that touch styling. Emits/consumes the single `.design/` artifact contract and scores on the canonical DESIGN-EVAL v1 rubric (9 dims, 0-4) — the legacy 10-dim 0-10 visual-audit scale is RETIRED.
design-sync
Use when design sync is needed. Push a local component library to a claude.ai/design design-system project incrementally — one component at a time, never a wholesale replace. Builds self-contained @dsCard preview HTML bundles, validates the card manifest, then drives the DesignSync MCP push flow.
design-studio
ONE entry to the whole design pipeline over a running app (Reframe/momentstudio/WebCoder/bilbo), the mode-routed successor to design-review-loop. Runs the full TCR design-review loop (8 gates G0-G7, revert-gate) plus a mode router: direct, audit, mirror, critique, iterate, implement, verify, loop, drive, session. NOT for a design DOCUMENT (metaswarm review-design) nor rubric math (ui-audit).
design-studio-session
The interactive-modify capability behind `design-studio session` mode. Use when the owner wants to point at a LIVE element in a running app (Reframe, momentstudio, WebCoder, bilbo) and iteratively restyle it with Claude and Codex — preview candidates as injected style overlays FIRST, resolve the clicked element back to its source file:line, apply ONE scoped micro-commit, incrementally re-audit only the changed screen, and auto-decide KEEP or REVERT through the SAME revert-gate as the batch design-review loop. Do NOT use for a whole-app batch audit (use design-studio / ui-audit) or for driving the app without modifying it (use ui-drive).
design-review-loop
MOVED - renamed to design-studio. Redirect tombstone kept so old invocations still resolve. Use design-studio instead; it is the SAME TCR design-review loop (G0 PRIME to G7 VERIFY, revert-gate) plus a mode router (direct, audit, mirror, critique, iterate, implement, verify, loop, drive, session). Old triggers - design review, TCR loop, mirror to Figma - route to design-studio.
delegated-build-verify
Use when delegating a large build to a generator (Codex fan-out, subagents, or a Workflow) and you must independently verify the result - the "agent builds X, then you evaluate" pattern. Turns a converged plan/grill into an authoritative build-spec that doubles as the evaluation rubric, fans out the generator, then verifies with an active fabrication hunt. Triggers: "build X then evaluate".
deepsource
Use when asked to retrieve DeepSource code review results — issues, vulnerabilities, report cards, coverage metrics, or analysis status ; do NOT use for running new scans outside DeepSource ; returns DeepSource findings and analysis run data.
deep-research
Multi-source deep research with a routing + caching + grading harness over firecrawl, exa, tavily, perplexity, scholarly-access, tooluniverse, huggingface, and context7 MCPs. Fixes scope first via AskUserQuestion (purpose, depth, recency, output form) before any fan-out, then routes each sub-question to the best live provider, caches provider calls by query hash, grades every source by confidence, and delivers cited reports. Use when the user wants thorough, evidence-backed research.
ddi-admet-research
Use as the front-end/router for biomedical research via the tooluniverse MCP server (Harvard Zitnik lab, 2206 tools) - drug-drug interactions, ADMET/physchem prediction, CYP450 enzyme roles, target research, FDA drug labels, drug repurposing, and pharmacovigilance. Routes to the detailed tooluniverse-* skills. Use when a question needs biomedical knowledge-base lookups.
data-scraper-agent
Use when building an automated public-data collector that scrapes on a schedule, enriches records with an LLM, stores results, and learns from feedback.