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 196852
format-codex-sessions
Use when Codex rollout JSONL files need to be turned into readable Markdown conversations, transcripts, or chat logs without dumping raw event noise, prompt bulk, or secrets.
forge-discipline
Use when turning a named behavioral pattern (gem, fresh arXiv finding, matured instinct) into an INSTALLED, self-tested Claude Code hook in one pass. A BEHAVIOR forge, not a tool forge - it forges hooks/tripwires the harness runs deterministically, not MCP tools. Scaffolds the hook, runs a chaos self-test proving it fires and fails open on garbage, then stages it with the one-line install.
firecrawl
Web scraping, crawling, structured extraction, site mapping, and change monitoring via the Firecrawl MCP. Use to turn URLs into clean markdown or schema-structured JSON, extract brand identity (colors, fonts, spacing), crawl or map sites, parse local PDFs/Office files, search the web with operators, or monitor pages for changes. Pairs with the Exa MCP (Exa finds by meaning, Firecrawl fetches).
finishing-a-development-branch
Use when implementation is complete, all tests pass, and you need to decide how to integrate the work - guides completion of development work by presenting structured options for merge, PR, or cleanup
figma
Use when fetching Figma design context, screenshots, variables, or assets via the Figma MCP server, or when setting up and troubleshooting the Figma MCP connection ; do NOT use for pixel-perfect full-component implementation from Figma nodes (use figma-implement-design) ; returns Figma design data and MCP setup guidance.
figma-implement-design
Use when the user provides Figma URLs or node IDs and needs production-ready code with pixel-perfect 1:1 fidelity to the design, translating Figma components into project-convention code via the Figma MCP ; do NOT use for generic Figma data fetching or MCP setup (use figma) ; returns implemented component code matching Figma specs.
fal-ai-media
Unified media generation via fal.ai MCP — image, video, and audio. Covers text-to-image (Nano Banana), text/image-to-video (Seedance, Kling, Veo 3), text-to-speech (CSM-1B), and video-to-audio (ThinkSound). Use when the user wants to generate images, videos, or audio with AI.
faculty-course-design-evidence
Use when designing or revising a course, syllabus, or module with evidence-based learning structure and assessment alignment.
fable-safe-prompt
Rewrite allowed prompts to reduce false-positive safety triggers without bypassing policy or changing intent.
execution-grounded-bon
Use when you have SEVERAL candidate patches/solutions for the same task — typically from a cross-vendor Claude+Codex worktree fan-out — and must pick the best one WITHOUT trusting an LLM to judge them (JudgeBench shows model-opinion rerank ≈ chance). Synthesize distinguishing inputs where the candidates diverge, execute all candidates on them, cluster by behavior (behavioral-diff), knock out crashers/timeouts, and select by EXECUTION-CONSENSUS — where an independent Claude candidate and an independent Codex candidate agree, that agreement is the signal (cross-vendor = independent failures). A BoN-MAV aspect jury ranks within the winning cluster. Distinct from A3 fail-first (validates ONE author's own tests) and diff-blind-test-synth (one patch vs spec-derived tests). Ships SHADOW: observe + log only, never blocks. Evidence: MASTER A4, COUNCIL C3; arXiv 2502.14382 (S*), 2409.08692 (B4), 2503.15223 (PatchDiff), 2207.10397 (CodeT).
executing-plans
Use when loading a written implementation plan and executing tasks in batches with user review checkpoints between batches, typically in a separate session ; do NOT use when all tasks are independent and can be dispatched in parallel within the current session (use subagent-driven-development) ; returns batch-by-batch progress with review pauses.
evidence-ledger-synthesis
Use when the user needs a claim-by-claim evidence ledger before writing a review, memo, paper section, or recommendation.
eval-harness
Formal evaluation framework for Claude Code sessions implementing eval-driven development (EDD) principles
estate-sync
Bootstrap, reconcile and audit this estate across machines. Use when setting up a second machine from the private repo, when the VM and home have drifted, when the VM/home file link is down, or when checking whether committed work was ever actually pushed. Also carries the private-estate plaintext credential policy.
estate-offensive-doctrine
FULL estate offensive-security doctrine (tiered out of the always-injected HOME AGENTS.md on 2026-09-21 — operator-approved context-tiering). READ THIS SKILL before any security assessment, engagement work, authorized target research, credential-store stewardship, or access-integrity research on the owner estate. Same rules as always — this file just isn't auto-injected into every request.
e2e-testing
Playwright E2E patterns -- Page Object Model, config, CI/CD, artifacts, flaky tests. NOT for driving a live app (use webapp-testing).
dump-to-signature-workflow
Use when converting dump or debugger evidence into durable signatures, AOB patterns, or diffable artifacts, especially when findings must survive across builds or runtime sessions.
drug-discovery-literature-bridge
Use when extracting molecules, targets, assays, and clinical context from a paper (via the Scholarly Access MCP paper_* tools) to seed the drug-discovery stack — ToolUniverse DDI/ADMET/target-validation/mechanism tools and the RDKit MCP — keeping DOI provenance on every extracted entity.