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
visual-explainer
Use when a visual explanation beats prose — diagrams, architecture overviews, diff or plan reviews, project recaps, comparison tables, dashboards, timelines, or HTML slide decks. Generates one self-contained HTML page (embedded CSS/JS, interactive Mermaid) that explains systems, code changes, plans, data, or technical concepts.
video-watermark-removal
Use when removing a static logo, watermark, or TV bug from a video seamlessly. PRIMARY method = ProPainter flow-guided temporal inpaint on a TIGHT logo-only mask (pulls REAL background pixels from neighbouring frames -> sharp, no soft haze), CHUNKED in time to fit a small GPU, with PER-FRAME presence gating so occluded frames stay pristine. big-LAMA is the fast SOFT fallback.
video-stabilize
Use when stabilizing a shaky / handheld / camera-shake video so it looks steadied (locked-down or smooth-pan). PRIMARY method = ffmpeg libvidstab TWO-PASS — vidstabdetect (analyze → writes a .trf transforms file) then vidstabtransform (read the .trf → render the steadied clip), with adaptive.
video-resize-aspect
Use when changing a video ASPECT RATIO or RESOLUTION for social delivery (9:16 / 1:1 / 4:5 / 16:9) - horizontal into a vertical Reel/Short, squaring, or scaling to a canvas, with scale + crop and SAR/DAR math. NOT for subject-tracking reframe (use reframe) or codec changes (use video-convert).
video-join-concat
Use when joining/concatenating multiple clips into ONE output, or assembling approved clips into an editable NLE timeline (CMX3600 EDL / CSV for Premiere/Resolve). Covers both media_studio ffmpeg paths, concat demuxer and filter. NOT for re-encoding one clip (use video-convert).
video-convert
Use when converting a video/audio file's FORMAT, CODEC, or CONTAINER in media-studio (mp4/mkv/mov/webm/gif, h264/h265/vp9/av1, or audio-only extraction to mp3/m4a/wav/flac/opus) — single file or batch, or normalizing a non-Chromium-playable source for playback (remux vs h264-720p proxy). Covers the real ffmpeg argv builder, codec/CRF/bitrate/pix_fmt choices, faststart, container<->codec legality, and the progress/seek/timebase pitfalls that actually bite. DO NOT use this skill for: cutting/trimming/splitting clips (use reframe-video-cutting), reframing 16:9->9:16 (use reframe), burning subtitles/captions, watermark/logo removal (use video-watermark-removal), or pure thumbnail/frame extraction. Also NOT a general 'run any ffmpeg command' helper — it documents media-studio's specific, frozen convert.* / media.* contracts.
verification-loop
A comprehensive verification system for Claude Code sessions.
verification-before-completion
Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always
vercel-deploy
Use when deploying an application or website to Vercel, creating preview deployments, or getting a Vercel deployment link ; do NOT use for Cloudflare Workers/Pages (use cloudflare-deploy), Netlify (use netlify-deploy), or Render (use render-deploy) ; returns Vercel deployment URL.
using-superpowers
Use when starting any conversation - establishes how to find and use skills, requiring Skill tool invocation before ANY response including clarifying questions
humanize
Rewrite AI-sounding text so it reads like the writer without changing what it says. Use when editing or reviewing prose for AI tells: not-X-but-Y contrasts, one-line closers, staged openers, forced triads, dashes everywhere, inflated claims, sales language, stock AI words, bold labels, or filler. Based on Wikipedia's "Signs of AI writing."
using-git-worktrees
Use when feature work needs isolation from the current workspace, or before executing a plan - creates isolated worktrees with safety checks. NOT for parallel agents (use dispatching-parallel-agents).
use-image-clipboard
Use when needing to read an image from the Windows clipboard and save it to a temporary PNG file on disk, returning its path for downstream use in scripts or pipelines. Required workaround for Antigravity CLI, which does not support direct image paste (Alt+V) the way Claude Code and Codex CLI do.
ui-ux-pro-max
Use when designing UI/UX for web or mobile apps and needing opinionated style, color palette, font pairing, or chart-type recommendations from a searchable database of 67 styles, 96 palettes, 57 font pairings, and 99 UX guidelines across 13 stacks ; do NOT use for code implementation or pixel-perfect Figma work ; returns ranked design recommendations.
ui-invariant-walk
Use when you want a property-based random-walk oracle that DRIVES a running web/Electron app and asserts UI invariants over a randomized command sequence — "no enabled button is a dead click" (an enabled, visible control must produce an observable state change) and "no route errors" (no navigation may reach an error route, throw an uncaught exception, or log a console error). Model-based testing via fast-check (`fc.commands` + `fc.asyncModelRun`) driving Playwright, with automatic shrinking to a minimal reproducing click/nav sequence. Ships as a SHADOW-mode advisory oracle (observe + log to a bounded rotating sink, never blocks CI or nags). Optionally deepen the existing Python Hypothesis suite with HypoFuzz nightly — but ONLY after verifying HypoFuzz's shifted license; the fast-check oracle ships regardless. NOT for pixel/design critique (use ui-audit), NOT for deterministic single-flow E2E (use Playwright directly).
ui-drive
Thin rail-router driver for a LIVE running web or Electron app — attach to it, resolve any target through three fallback rails (a11y-ref then CDP css then vision xy), point at an element with a screenshot-visible highlight, discuss it with full context, and actuate it (click/fill/hover/press/select). Also drives the ephemeral restyle loop — preview a CSS candidate as an injected style overlay, apply/accept/revert it, and harvest the accepted deltas as a PROPOSED patch — WITHOUT ever editing application source. Rides the vendored playwright at exactly 1.61.1 from the sibling ui-audit skill, so no new browser dependency. Every mutating verb logs to a bounded shadow sink (observe-only, never blocks). Use as the low-level app-driving and pointing channel under design-studio-session and design-studio. NOT for the design-quality verdict (use ui-audit) and NOT for element-to-source mapping or micro-commits (use design-studio-session).
ui-audit
Drive a running web or Electron app and emit a deterministic UI/UX + functional audit - a11y, contrast, spacing, tap targets, dead buttons, motion, visual regression (Track-A) - then a design-quality VERDICT (Track-B) where Claude reasons over the numbers + screenshot on the DESIGN-EVAL v1 rubric (0-4, critical-error override, weighted composite) against Refactoring-UI/HIG/Nielsen/WCAG. READ-ONLY.
two-tier-verify
Use when repeated code edits are triggering a full test-suite rerun each iteration and burning tokens/time, or before any commit. Splits verification into a cheap INNER loop per edit (only the affected/related tests plus type-check and lint on touched files) and a single OUTER loop at the commit boundary (full suite plus coverage plus held-out E2E/golden-journey plus a scoped mutation check). Gives concrete jest/vitest/playwright and pytest command recipes and the decision rule for which tier to run. NOT a replacement for the golden-journey done-signal (see seed-golden-journey); NOT for one-off single-file scripts where a full run is already trivial.