Claude Skills
The open format is called Agent Skills and works in Claude Code, Codex, Cursor and other agents — most people know it as Claude Skills.
Every Agent Skill we could find on GitHub, deduplicated by content. 86 961 files from 2 125 authors, of which 68 678 are unique — the rest is the same skill repackaged into someone else's repository. For each one: what it weighs in tokens, whether it ships runnable scripts, and which MCP servers it needs.
Use when deploying changes to staging across relay, relay-dashboard, and relay-cloud repos - coordinates multi-repo branch syncing using git worktrees, automatically triggers staging deployments via GitHub Actions
Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.
Use when implementing GitHub OAuth + GitHub App authentication with Nango - provides two-connection pattern for user login and repo access with webhook handling
Use when building Cmd+K command palettes in React - covers keyboard navigation with arrow keys, keeping selected items in view with scrollIntoView, filtering with shortcut matching, and preventing infinite re-renders from reference instability
Best practices for structuring prpm.json package manifests with required fields, tags, organization, multi-package management, enhanced file format, eager/lazy activation, and conversion hints
Use when writing agent-relay workflows that must fully validate features end-to-end before merging. Covers the 80-to-100 pattern - going beyond "code compiles" to "feature works, tested E2E locally." Includes repair-before-failure validation gates, mandatory sequential Claude-then-Codex fresh-eyes review/fix loops with test hardening, PGlite for in-memory Postgres testing, mock sandbox patterns, test-fix-rerun loops, verify gates after every edit, and the full lifecycle from implementation through passing tests to commit.
Use when writing Agent Relay or Ricky workflows that must loop review, fix, and validation with fresh agent context until independent signoff agents, typically Claude and Codex, both agree the work is comprehensively complete. Covers fresh-context iterations, repairable gates, dual reviewer verdict contracts, iteration-count reporting, PR signoff comments, and blocked-state handling.
Use when an agent or human needs to set up relayfile end-to-end so agents can read and write provider files through a local mount. Covers `relayfile setup`, dynamic integration discovery with `relayfile integration available/search`, Nango and Composio backend selection, Atlassian site selection and metadata, cloud login, OAuth/connect flows, mount verification, `RELAYFILE_LOCAL_DIR` handoff, writeback status and retry commands, and key May 2026 cloud-mount gotchas.
Use when building multi-agent workflows with @relayflows/core. Covers conversation vs pipeline coordination, WorkflowBuilder/DAG steps, agents, {{steps.X.output}} chaining, repairable verification gates, evidence-based completion, mandatory Claude-then-Codex fresh-eyes review/fix loops with test hardening, channels, chat-native recipes, error handling, event listeners, step sizing, lead+workers teams, and parallel waves.
Use when you need Codex to coordinate multiple agents through Agent Relay for peer-to-peer messaging, lead/worker handoffs, or shared status tracking across sub-agents and terminals.
Meta-skill that generates new design language skills. Works on Claude Code and Codex. Use when the user says 'create a design skill', 'generate design language', 'new design system skill', 'design skill inspired by X', 'design skill from this screenshot', '/hue', or 'use hue'. Also triggers for 'remix my design skill' or 'make my skill more X'.
Analyzes recently modified code and creates pull requests with simplifications that improve clarity, consistency, and maintainability while preserving functionality
Validate KQL (Kusto Query Language) files used in Azure Quick Review (azqr) against their recommendation definitions. Use when the user wants to validate KQL syntax, check semantic alignment with recommendations, verify property names against Azure REST API schemas, or audit KQL queries before a pull request. WHEN: "validate kql", "check kql files", "kql syntax error", "validate aks kql", "run kql validator", "validate azure resource graph queries", "check recommendations alignment", "validate kql for <service>".
Expert guidance for developing and contributing to Azure Quick Review (azqr) - A Go-based CLI tool for Azure resource compliance analysis
Monsters, dragons, world reveals. Each entity gets its own definition block, the shots are cut by timecode, and scale is bought with reference objects and low angles rather than words like massive. Use when the user wants a Seedance 2.5 / 2.0 video prompt for this kind of clip (Epic fantasy and sci-fi spectacle; 中文触发词: 奇幻科幻大场面, 奇幻, 科幻, 巨龙, 怪兽, 史诗), or asks how published cases of this type were prompted. Read references/cases.md before drafting.
One small anthropomorphic character carries the whole film. Spell out its looks part by part, write the style as render settings you can measure, and cut the timeline so each block holds a single action goal. Use when the user wants a Seedance 2.5 / 2.0 video prompt for this kind of clip (3D cartoon character short; 中文触发词: 3D 卡通角色短片, 3D 卡通, 皮克斯, 黏土, 卡通角色), or asks how published cases of this type were prompted. Read references/cases.md before drafting.
Verified prompt templates and structures for Seedance 2.5 / 2.0 video generation, distilled from the highest-performing published cases on goodcase.ai. Use when the user asks to write, revise, or review a Seedance video prompt, or asks how to prompt Seedance for a specific kind of shot (narrative short, ad/commercial, UGC talk-to-camera, dialogue with lip-sync, action, product reveal, retro home video, travel film, pet or animal lead, comedy skit, horror, fantasy or sci-fi spectacle, 3D cartoon, car spot, fashion film, sports, storyboard image to video, and similar). Also use when the user names Seedance, 即梦视频, or Seedance 2.5/2.0 directly.
The machine has to stay one machine while the camera does all the work. Lock the vehicle part by part, then fill the runtime with a numbered cut list that moves the camera every second. Use when the user wants a Seedance 2.5 / 2.0 video prompt for this kind of clip (Cars and vehicles at speed; 中文触发词: 汽车与载具速度片, 汽车, 赛车, 摩托, 载具), or asks how published cases of this type were prompted. Read references/cases.md before drafting.
A short skit where every beat is laid down to serve one punchline. The joke has to land on something visible, and the absurdity only works when the camera and the physics stay dead serious. Use when the user wants a Seedance 2.5 / 2.0 video prompt for this kind of clip (Twist-ending comedy skit; 中文触发词: 反转结尾搞笑短片, 搞笑, 反转, 梗, 沙雕), or asks how published cases of this type were prompted. Read references/cases.md before drafting.
Every shot carries its own timecode and shows one visible change on a body. The dread comes from the chain — a look, veins, a bite, the next person — and the ending seals a door without settling anything. Use when the user wants a Seedance 2.5 / 2.0 video prompt for this kind of clip (Horror and suspense; 中文触发词: 恐怖悬疑短片, 恐怖, 悬疑, 惊悚), or asks how published cases of this type were prompted. Read references/cases.md before drafting.
One person, one look, a handful of places. A head-to-toe appearance lock carries the whole clip, and each scene gets one location, one gesture and one kind of light. Use when the user wants a Seedance 2.5 / 2.0 video prompt for this kind of clip (Fashion lookbook and portrait film; 中文触发词: 时尚 lookbook 与人像写真片, 时尚, 穿搭, 人像, 写真), or asks how published cases of this type were prompted. Read references/cases.md before drafting.
The animal is the lead and a phone is the only camera. Lock the count to exactly one, keep the animal behaving like an animal, and let the payoff come from it closing in on the lens. Use when the user wants a Seedance 2.5 / 2.0 video prompt for this kind of clip (Pets and animals as the lead; 中文触发词: 宠物动物当主角, 宠物, 猫, 狗, 动物), or asks how published cases of this type were prompted. Read references/cases.md before drafting.
The clip lives or dies on the action loop. Write every link from run-up to landing in order, name the physics you want by part, and spend the negative list on flying, hovering and teleporting. Use when the user wants a Seedance 2.5 / 2.0 video prompt for this kind of clip (Sports and extreme stunts; 中文触发词: 体育与极限运动, 体育, 极限运动, 跑酷, 滑板), or asks how published cases of this type were prompted. Read references/cases.md before drafting.
Two stages: first a single-page sheet of numbered panels from an image model, then that sheet fed to Seedance as the reference. The sheet owns order, framing and timing, and the video prompt only has to connect the panels. Use when the user wants a Seedance 2.5 / 2.0 video prompt for this kind of clip (Storyboard grid to video; 中文触发词: 分镜网格转视频, 分镜图, 九宫格, 故事板), or asks how published cases of this type were prompted. Read references/cases.md before drafting.
The period feel comes from camera defects and one tiny everyday incident. Write the camcorder's autofocus hunting, exposure shifts and handheld shake as an explicit list, keep the story small, and the clip reads as a real old tape. Use when the user wants a Seedance 2.5 / 2.0 video prompt for this kind of clip (Early-2000s DV home video; 中文触发词: 早年 DV 家庭录像, DV, 家庭录像, VHS, 复古, 年代感), or asks how published cases of this type were prompted. Read references/cases.md before drafting.
One traveller moves through a place scene by scene, each scene carrying its own timecode, its own location and one short line. The polish is bought with film grain and golden-hour light. Use when the user wants a Seedance 2.5 / 2.0 video prompt for this kind of clip (Cinematic travel vlog montage; 中文触发词: 电影感旅行漫游, 旅拍, 城市漫游, 旅行), or asks how published cases of this type were prompted. Read references/cases.md before drafting.
Audit the Neovim plugins in this dotfiles repo one by one. Researches what current Neovim already covers and each plugin's maintenance status and news, asks the user to keep, replace or remove each one with AskUserQuestion, then applies the decisions (keymaps rerouted, :Lazy clean, README), verifies with headless Neovim and commits. Use for "audit my Neovim plugins", "棚卸しして", "Neovim のプラグインを整理したい".
Connect an MCP-capable coding agent to OpenChatCut and edit local video projects. Use when the user asks to install, connect, or set up OpenChatCut; inspect or edit an OpenChatCut project; work with its timeline, transcript, captions, media, generation, motion graphics, audio, color, or export tools; or recover from an OpenChatCut MCP error.
Plan AI short films with story, shots, prompts, and continuity.
Use when acquiring or importing media into a OpenChatCut project asset library for video editing or creation, including local/attached videos, user-provided paths, public media URLs, web video/audio/image assets, upload fallback decisions, and deciding between import_media, download_media, or manual user action.
Use whenever the agent needs to add, create, hand-author, patch, or place Motion Graphic JSX assets in a OpenChatCut project. This is the direct-authoring path: use create_motion_graphic_from_code / edit_asset / edit_item, not motion-graphic-gen or submit_motion_graphic. Covers project/timeline intake, project visual language, editable properties, asset binding, inline JSX authoring, existing asset updates, timeline placement, and verification.
Create finished explainer videos from a topic, script, outline, voiceover, product logic, data, technical concept, course material, or reference assets. Use when the user wants narration, motion graphics, stock footage, generated visuals, or mixed visuals to explain an idea.
Use when a OpenChatCut video editing or creation workflow needs export, render, download, share, final delivery, subtitle-file export, render choice, local-only asset handling, or export fallback explanation.
| AI image generation via gpt-image-2, nano-banana, and MiniMax image-01. Use when the user wants to generate or create an image / picture / still.
Use when a OpenChatCut tool call fails or returns an unexpected shape.
Cut one long podcast, interview, course, livestream, or other source video into social-ready shorts, reels, highlights, or clip timelines from existing project media. Use when the user asks to cut a long video into Shorts, Reels, TikToks, Xiaohongshu posts, best moments, highlights, or multiple clips.
Add motion graphics at the right moments without blocking the story. Use when the user wants to enhance talking-head, lecture, tutorial, interview, podcast, or creator videos with motion graphics.
Turn multiple product shots, event footage, travel clips, gameplay moments, UGC/product footage, B-roll, or mixed media into social-ready reels, highlights, recaps, or montage-style short videos from existing project media.
| Music generation via Mureka and MiniMax. Use for instrumentals, songs, soundtracks, track/stem generation, or covers through `submit_music`.
Use for video editing or video creation work that should be editable in OpenChatCut, even when the user does not explicitly mention OpenChatCut. Covers local/attached video editing, captions/subtitles, transcription, trimming, talking-head cleanup, highlights, B-roll, overlays, generation, export, project/editor opening, importing, targeting, verifying, watching, and identifying the active OpenChatCut project/editor URL.
Turn a product into ad angles, hooks, scenes, and CTA.
| OpenChatCut product knowledge — UI layout, editor features, and how generation providers are configured. Use when the user asks about the product interface, how to use a feature, where to find something, or needs GUI guidance for something the agent cannot do directly. Also use as fallback when a task fails and the user needs to complete it manually in the UI. NOT for live project-state queries ("where are my folders?", "what's on my timeline?", "where is clip X?") — those are answered by `read_project`, not by this skill.
| AI shader generator for WebGL video effects, transitions, masks, and color grading (LUT / 调色 / 电影感 / film look). Use when the user wants a video effect (滤镜 / 特效), a transition (转场 / crossfade / wipe / cube / 3d), a mask (蒙版 / 遮罩 / reveal), a zoom / push-in (推近 / 推镜头), or a color grade — try the built-in effects (zoom, builtin LUTs) before generating a new shader.
Break down each shot and turn the analysis into a storyboard reference. Use when the user wants shot-by-shot film analysis, director logic, cinematography breakdown, or a storyboard-style reference from a video.
| Guide for editing videos where the primary content is people talking — talking-head / 口播, interview / 访谈, lecture, tutorial, podcast, course content, and similar talking-driven formats. Use when the user wants speech editing on a talking video (剪口播 / 口播剪辑 / 去口癖 / clean up fillers / smooth speech), motion graphics layered onto talking video (口播加 MG / 加动画), or B-roll on a talking video (加 B-roll / add B-roll). For motion graphics specifically, use this together with the active Motion Graphics skill/workflow available in the current OpenChatCut environment — this skill adds talking-specific guidance (speech-rhythm timing, frame-aware placement, subject/caption protection, placement verification).
Use when a video/audio task needs OpenChatCut transcription, captions, subtitles, subtitle styling, transcript search, transcript readiness checks, or enabling captions, including local or attached videos where the user asks to add captions/subtitles, transcribe, create bilingual subtitles, clean talking-head speech, remove filler words, or trim pauses.
Use when checking whether agent edits are reflected in the OpenChatCut project and editor.
| AI video generation via Seedance 2.0, Kling, and MiniMax Hailuo. Use when the user wants to generate a video clip — text-to-video, image-to-video, first/last-frame transitions, reference-guided generation, multi-shot, or generatively editing / extending an existing clip.
Create platform-ready thumbnails from real video frames. Use when the user wants a thumbnail, cover image, YouTube cover, Shorts cover, Bilibili cover, Xiaohongshu cover, or other video poster image.
| Text-to-Speech (TTS), voiceover, narration placement/sync, and custom sound effects (SFX) generator. Use when the user wants generated speech from text, wants to add/replace/align narration or voiceover for an existing video/timeline, wants to keep existing voiceover synced after visual retiming edits, needs voice audition/selection, or explicitly wants a newly generated/custom sound effect that is not available in the Sound Effects library.
Use when the agent should ask the user for structured input with an in-chat form, including single-select, multi-select, text fields, style pickers, or voice audition choices.
Build Amazon Machine Images (AMIs) with Packer using the amazon-ebs builder. Use when creating custom AMIs for EC2 instances.
Build Azure managed images and Azure Compute Gallery images with Packer. Use when creating custom images for Azure VMs.
Build Windows images with Packer using WinRM communicator and PowerShell provisioners. Use when creating Windows AMIs, Azure images, or VMware templates.
Push Packer build metadata to HCP Packer registry for tracking and managing image lifecycle. Use when integrating Packer builds with HCP Packer for version control and governance.
Azure Verified Modules (AVM) requirements and best practices for developing certified Azure Terraform modules. Use when creating or reviewing Azure modules that need AVM certification.
Discover existing cloud resources using Terraform Search queries and bulk import them into Terraform management. Use when bringing unmanaged infrastructure under Terraform control, auditing cloud resources, or migrating to IaC.
Generate Terraform HCL code following HashiCorp's official style conventions and best practices. Use when writing, reviewing, or generating Terraform configurations.
Comprehensive guide for writing and running Terraform tests. Use when creating test files (.tftest.hcl), writing test scenarios with run blocks, validating infrastructure behavior with assertions, mocking providers and data sources, testing module outputs and resource configurations, or troubleshooting Terraform test syntax and execution.
Transform monolithic Terraform configurations into reusable, maintainable modules following HashiCorp's module design principles and community best practices.
Claude Skills — questions
Answers built from the skills we actually parsed.