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.
Initialize evo for the current repository by exploring the codebase, proposing unexplored optimization dimensions, constructing the benchmark inside a baseline worktree, and running the first experiment. Use when the user invokes /evo:discover, mentions setting up evo, wants to instrument a codebase for autonomous optimization, or asks to start a new evo run on a project.
Non-user-invocable provider/setup reference for evo backend switching, prerequisite checks, and auth/install guidance.
Drive structured autoresearch iteration after evo:discover and the baseline commit. Use when the user invokes /evo:optimize or asks to try ideas, try variants, run experiments, use available GPUs, improve the current best/frontier, continue an evo search, or compare candidate changes in an evo workspace. The orchestrator plans and spawns optimization subagents; candidate edits/runs belong to those subagents. Width is set via subagents=N (1 for serial workloads, larger for parallel); the loop's structural value applies at any width.
Read-only evo run reporting. Use when the user invokes /evo:report, asks what happened overnight, asks what improved recently, asks for the best/frontier candidates, asks for a quick score chart without opening the dashboard, or wants the scatter plot in chat output. Never run benchmarks, gates, Slurm commands, evo run, or ad-hoc verification scripts for report requests.
Land the winning experiment from an evo run as a clean, mergeable change -- open a PR when the repo has a remote, otherwise merge into the working branch. Distills the best-scoring experiment down to the minimal diff that reproduces its behaviour, shaped for the qualities a maintainer merges on (scope discipline, test integrity, style adherence), then attaches an advisory mergeability report. Use when the user invokes /evo:ship, asks to land/merge/ship the best result, or wants to turn a finished optimization into a pull request.
Protocol that evo optimization subagents follow when dispatched from /optimize. Auto-loaded by spawned subagents via their host's skill loader. The orchestrator may also invoke this skill to understand the brief shape its dispatched subagents expect + what they're required to emit -- useful when writing briefs or debugging a subagent's behavior.
This skill should be used when picking or diagnosing a training move (SFT, LoRA, DPO/KTO/ORPO, RFT, GRPO/PPO/RLOO, RLHF), or when the user mentions fine-tuning, post-training, training recipe, reward design, or weight updates. Decision tree by reward shape, smoke-run gate, three failure diagnostics, five false-progress patterns. Provider recipes and I/O contract in references/.
Run the evo optimization loop with parallel subagents until interrupted.
Plan and implement OpenMeld PostgreSQL business persistence, schemas and migrations, transactions, Hyperdrive access, and data cutover. Use for the shared relational database transition or PostgreSQL work; not for unrelated D1 performance fixes or OpenConnector's upstream database.
Maintain the official OpenMeld Changelog from repository evidence. Use when deciding whether a change needs a Changelog entry, writing or publishing an entry, preparing release notes, validating generated Changelog files, or historically backfilling missing changes.
> Debugs failing GitHub Actions CI runs for PostHog PRs, commits, and branches, and answers broad CI-health questions ("is CI red?", "is master green today?", "what's broken right now?"). Use when the user asks why CI is red, asks for the current CI or master status, or mentions a failing check, GitHub Actions run, Depot runner, workflow, job, shard, merge queue kick, flaky test, lint failure, typecheck failure, snapshot diff, migration check, generated types drift, or skills from engineering analytics), then guides read-only inspection, failure classification, smallest local reproduction with hogli, and safe reporting without rerunning CI or posting to GitHub. Running unattended as the
Use when closing out a feature, tidying developing/ documentation, or asked to check docs health — runs the lifecycle linter and applies the README deletion rules.
Use when deciding whether an OpenMeld change needs a Changelog entry, writing or publishing an entry, preparing release notes, validating Changelog projections, or historically backfilling missing changes.
> Guides an agent through reproducing, root-causing, fixing, and validating flaky tests in the PostHog monorepo. Stabilizing is not the only valid outcome — the skill also gates whether the test should exist, so deleting a test that catches nothing real, or re-leveling one that flakes because of the level it runs at, are first-class endings.
Use when the user requests React Doctor or a React code-health audit, or when a concrete React diagnostic question would benefit from its scan. Not a default step after React changes or before committing.
> Gates whether a new test should exist and forces it to be efficient, protecting CI from low-value test bloat. Use before any change to what a pytest, Jest, or Playwright test asserts or sets up, down to one fixture or one assertion added to an existing block. Front-loads the value bar (every test must catch a realistic regression no existing test already catches; extend the nearest existing test before writing a new standalone one; test behavior through the public interface, not implementation details; collapse near-duplicates into parameterized cases) and the efficiency bar (deterministic, isolated, fast; pick the cheapest test level; Django TestCase over TransactionTestCase; no sleeps, no real network; no time bombs from absolute dates left to age against the real clock; no database a test never uses). Includes a "don't write it" decision tree. For fixing an existing flaky test use `/fixing-flaky-tests`; after this gate says a Playwright test is warranted, use `/playwright-test` for mechanics.
Use when asked to deploy, release, publish, or ship OpenMeld to production — workers, Web, CLI, or local components. Routes to the guarded release lanes; never deploy raw.
Official public OpenMeld Skill for agents using the `openmeld` CLI to consult the Center Agent, set up OpenMeld, connect local agents, create or request Agent Profiles, work in Organizations and Spaces, observe or stop Wakes, trace delivery, and repair local state.
Use OpenMeld through its public CLI to work in Spaces, reply to a Wake, consult the Center Agent, connect this computer, manage Agent Profiles, or diagnose delivery.
Use when diagnosing a Wake that failed, stalled, or produced no visible reply — dispatch, setup handoff, or reply-path issues. Routes to the deterministic trace evidence path.
Use OpenMeld through its public CLI to work in Spaces, reply to a Wake, consult the Center Agent, connect this computer, manage Agent Profiles, or diagnose delivery.
Use OpenMeld through its public CLI to work in Spaces, reply to a Wake, consult the Center Agent, connect this computer, manage Agent Profiles, or diagnose delivery.
Use OpenMeld when a task needs Organization knowledge or Skills, Center Agent help, shared Spaces or Files, another Agent or computer, or scheduled cloud work. Also use for OpenMeld setup and Wake replies.
Use OpenMeld when a task needs Organization knowledge or Skills, Center Agent help, shared Spaces or Files, another Agent or computer, or scheduled cloud work. Also use for OpenMeld setup and Wake replies.
> Monitors task execution for skill improvement opportunities. Use this skill during ANY multi-step task, agentic workflow, or substantive work session where the agent is using tools and producing deliverables. It captures patterns, user corrections, workflow insights, and methodology worth preserving as reusable skills. Also triggers during post-task feedback discussions and when the user explicitly mentions skill observations, improvements, the observation log, skill taxonomy, or asks the agent to watch for skill opportunities. Also known this skill should be invoked at the start of every task-oriented session — if you are about to use tools to produce deliverables, invoke this skill first. For reliable activation, pair this description with a CLAUDE.md instruction or harness-level session-start hook (see Recommended Activation Setup) — description-level matching alone is not enforceable.
Google Calendar: Manage calendars and events.
A set of resources to help me write all kinds of internal communications, using the formats that my company likes to use. Claude should use this skill whenever asked to write some sort of internal communications (status reports, leadership updates, 3P updates, company newsletters, FAQs, incident reports, project updates, etc.).
Creating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems. Create original algorithmic art rather than copying existing artists' work to avoid copyright violations.
Fetch GitHub issues, spawn sub-agents to implement fixes and open PRs, then monitor and address PR review comments. Usage: /gh-issues [owner/repo] [--label bug] [--limit 5] [--milestone v1.0] [--assignee @me] [--fork user/repo] [--watch] [--interval 5] [--reviews-only] [--cron] [--dry-run] [--model glm-5] [--notify-channel -1002381931352]
Analyze weekly marketing campaign performance data across channels. Use when analyzing multi-channel digital marketing data to calculate funnel metrics (CTR, CVR) and compare to benchmarks, compute cost and revenue efficiency metrics (ROAS, CPA, Net Profit), or get budget reallocation recommendations based on performance rules.
CraftedWell brand guidelines for presentations and documents. Use this skill whenever creating or styling documents (docx, pdf) or presentations (pptx) for CraftedWell. Apply warm, artisanal aesthetic with chocolate/caramel color palette, Georgia headings, and Arial body text.
Analyze weekly marketing campaign performance data across channels. Use when analyzing multi-channel digital marketing data to calculate funnel metrics (CTR, CVR) and compare to benchmarks, compute cost and revenue efficiency metrics (ROAS, CPA, Net Profit), or get budget reallocation recommendations based on performance rules.
Generate educational practice questions from lecture notes to test student understanding. Use when users request practice questions, exam preparation materials, study guides, or assessment items based on lecture content.
Generate educational practice questions from lecture notes to test student understanding. Use when users request practice questions, exam preparation materials, study guides, or assessment items based on lecture content.
Provides Typer templates, handles registration, and ensures consistency. ALWAYS use this skill when adding or modifying CLI commands. Use when user requests to add/create/implement/build/write a new command (e.g., "add edit command", "create search feature") OR update/modify/change/edit an existing command.
Generate pytest tests for Typer CLI commands. Includes fixtures (temp_storage, sample_data), CliRunner patterns, confirmation handling (y/n/--force), and edge case coverage. Use when user asks to "write tests for", "test my CLI", "add test coverage", or any CLI + test request.
Provides checklist for reviewing Typer CLI command implementations. Covers structure, Annotated syntax, error handling, exit codes, display module usage, destructive action patterns, and help text conventions. Use when user asks to review/check/verify a CLI command, wants feedback on implementation, or asks if a command follows best practices.
Create learning paths for programming tools, and define what information should be researched to create learning guides. Use when user asks to learn, understand, or get started with any programming tool, library, or framework.
>- Diagnose, repair, and standardize repository setup and safe Git workflows for Claude Code or Codex. Use when a repository will not run, a collaborator is onboarding, dependencies or credentials are missing, the user wants startup sync, SessionStart output is duplicated, project instructions or hooks need auditing, or commit/push/conflict/history-cleanup needs a guarded workflow. Route ordinary startup behavior through project instructions or a natural language request; use lifecycle hooks only when behavior must occur before the first prompt and the target runtime has been verified.
> Runs adversarial due-diligence on a benchmark the user envies — a founder, KOL, company, or product whose claimed success looks inflated — splitting marketing bubble from real signal, then mapping the validated playbook onto the user's own resources. Use whenever the user wants to 尽调/对标/拆解 a competitor or role-model, 抄/偷师 someone's playbook, suspects 水分/泡沫 in their claims (#1 on Product Hunt, 0-to-1M users, funding, 估值几个亿), asks whether wins are 真本事 vs 运气/时机, or says someone is 太成功了/crushing it and wants the real story — even if they never say 尽调. Prefer over deep-research for debunking inflated claims and extracting a replicable playbook rather than a neutral briefing.
Fetch comprehensive, login-free data for any Bilibili (B站) video — title, UP name and follower count, publish date, partition, tags, per-part cids, live stats (view, like, coin, favorite, share, reply, danmaku), and full danmaku (bullet-comment) text. Use this skill whenever working with a Bilibili video and needing real, citable numbers or metadata — ingesting a Bilibili source into a knowledge base, analyzing why a video performed, verifying a creator's claimed metrics, building a case study, or any time a Bilibili view/like/favorite count is about to be written into a document — fetch it, never hand-type or estimate it. Accepts BVID, av numbers, b23.tv short links, or full URLs. Subtitles are also covered but require the user's Bilibili login.
Programmatic screenshot capture on macOS. Find window IDs with Swift CGWindowListCopyWindowInfo, control application windows via AppleScript (zoom, scroll, select), and capture with screencapture. Use when automating screenshots, capturing application windows for documentation, or building multi-shot visual workflows.
Generates professional animated CLI demos as GIFs using VHS terminal recordings. Handles tape file creation, self-bootstrapping demos with hidden setup, output noise filtering, post-processing speed-up, and frame-level verification. Use when users want to create terminal demos, record CLI workflows as GIFs, generate animated documentation, build demo tapes for README files, or need to showcase any command-line tool visually. Also triggers on "record terminal", "VHS tape", "demo GIF", "animate my CLI", or any request to visually demonstrate shell commands.
Investigate and resolve Cloudflare configuration issues using API-driven evidence gathering. Use when troubleshooting ERR_TOO_MANY_REDIRECTS, SSL errors, DNS issues, or any Cloudflare-related problems. Focus on systematic investigation using Cloudflare API to examine actual configuration rather than making assumptions.
Start or reuse a self-contained local web gallery for browsing Codex-generated images. Use when the user asks to browse Codex generated images, open a local image gallery, inspect ~/.codex/generated_images, view a Codex image output folder, or browse image files produced by Codex.
>- Discover, clone, update, and analyze competitor repositories with evidence-based competitive intelligence. Use when tracking competitors, reviewing competitor source code, adding a competitor repository, comparing product capabilities, building a competitor landscape, checking whether competitor code changed, or when the user says "竞品分析", "竞品", "competitor scan", "latest competitor code", "analyze competitor", or "compare with X". Repository-backed findings must come cite their source and volatility.
>-
> Transforms raw meeting transcripts into high-fidelity, structured meeting minutes (notes / summaries). Use when (1) a meeting transcript is provided and meeting minutes, notes, or a summary are requested; (2) multiple versions of minutes must be merged without losing content; (3) existing minutes need review against the original transcript for missing items; (4) the transcript has anonymous speakers like "Speaker 1/2/3" or "发言人1" that need identifying (optionally mapped via a context.md team directory). Triggers on 会议纪要 / 会议记录 / 整理纪要 / 妙记转纪要, "write meeting minutes", "summarize this meeting", "merge these minutes", "what's missing from these notes". For fixing ASR/STT recognition errors in the raw transcript first, use transcript-fixer; this skill structures clean transcripts into minutes.
Transcribe audio with StepFun's stepaudio-2.5-asr — an SSE endpoint (NOT /v1/audio/transcriptions) with 32K context, ~85-101x RTF on long audio, and a single-call ceiling around 30 minutes (no client-side chunking). Use when transcribing Chinese / English audio with StepFun, when long-form recordings (5-30 min) need to land in one request, when migrating from step-asr / step-asr-1.1, or when hitting the misleading `model stepaudio-2.5-asr not supported` error (which actually means wrong endpoint). Triggers on 阶跃 ASR, StepFun ASR, stepaudio-2.5-asr, 转录, 语音识别, 长音频转写, 语音转文字. For TTS with the sibling stepaudio-2.5-tts model, use the stepfun-tts skill instead.
Generate Chinese / Japanese speech with StepFun's stepaudio-2.5-tts — Contextual TTS that replaces step-tts-2's `voice_label` with natural-language `instruction` (≤200 chars) plus inline `()` parentheses for句内 prosody. Use when the user wants emotional / prosody control over voice synthesis (whisper, pause, stress, mood pivot mid-sentence), batch-generates game / app voice lines, migrates from `step-tts-2` (the `voice_label → instruction` breaking change), or hits StepFun's stricter 2.5-era censorship (死/消失/political terms). Triggers on 阶跃 TTS, StepAudio 合成, 语音合成, 配音, 文本转语音, TTS 升级, 迁移 step-tts-2. For transcription with the sibling stepaudio-2.5-asr model, use the stepfun-asr skill instead.
>- Corrects speech-to-text transcription errors using dictionary rules and Claude's built-in AI (no external API key required — Native AI Correction is the DEFAULT). Stage 3 API is a backup for automation without Claude Code. Builds personalized correction databases that learn from each fix, auto-loads person-name ASR variants from your people roster, and reads per-domain context files that prime the AI pass for context-dependent homophones. Triggers when working with ASR/STT output containing recognition errors, homophones, garbled technical terms, person-name errors, or Chinese/English mixed content. Also triggers on requests to clean up meeting notes, lecture transcripts, interview recordings, or any text produced by speech recognition. Use this skill even when the user just says "fix this transcript", "clean up these meeting notes", or mentions garbled names without invoking ASR specifically.
>- Searches and recovers Claude Code JSONL history across all active config homes and archives registered in ~/.claude/history-sources.json. Use --all-projects when the project is unknown and --codex to include Codex rollout search. Uses internal timestamps and searches messages, thinking, tool inputs/results, queues, attachments, summaries, titles, and file-history paths. Recovers exact captured bytes from Claude file-history snapshots, including post-Write edits and binary files; otherwise labels Write checkpoints as lower fidelity. Use for keyword/date-bounded history search, prior-conversation forensics, deleted-file recovery, vanished ~/.claude/jobs artifacts, tool/file-operation analysis, or requests mentioning session history, find in history, previous conversation, or .claude/projects. For a recent Claude+Codex inventory, use local-conversation-history instead.
>- How to write, test, register, and debug Claude Code hooks — PreToolUse / PostToolUse / SessionStart / Stop Bash guards that enforce a rule the model would otherwise talk itself past. Use whenever the user wants to create a hook, block/intercept a tool call, turn a repeatedly-violated rule into a hard gate, add a guard rail, debug a hook that misfires or "poisons the session", register a hook across profiles, or mentions hooks / PreToolUse / Stop hook / 拦截 / 守卫 / 钩子 / 拦下. Bakes in the hard-won text — a rule about Claude's own output belongs on Stop instead; token-level shlex matching (never awk splitting); bash -n + real-JSON end-to-end testing BEFORE registering (a corrupted PreToolUse hook poisons every Bash call); SSOT + symlink so a ~/.claude reinstall can't lose it; multi-profile convergence; and human-confirmation release gates. Reach for this even for "make it stop doing X" — a durable stop is a hook, not a reminder.
> Fixes broken line wrapping in Claude Code exported conversation files (.txt), reconstructing tables, paragraphs, paths, and tool calls that were hard-wrapped at fixed column widths. Includes an automated validation suite (generic, file-agnostic checks). Triggers when the user has a Claude Code export file with broken formatting, mentions "fix export", "fix conversation", "exported conversation", "make export readable", references a file matching YYYY-MM-DD-HHMMSS-*.txt, or has a .txt file with broken tables, split paths, or mangled tool output from Claude Code.
|
>- Migrates Claude Code personal memory (the per-project memory/ directory) into tool-agnostic reference docs, so other AI CLIs that auto-load AGENTS.md (Codex primarily; the content architecture transfers to Cursor and others) working in the same directory can read the same user profile, collaboration preferences, and methodology instead of being blind to them. Use this whenever the user says things like "migrate my memory", "my memory is locked to Claude Code", "make Codex/Cursor read my profile", "memory should live in docs not one tool", or reports that a second AI tool doesn't know who they are; also use it when memory has grown bloated with content that should be shared across tools or projects. Covers diagnosis, the references/ + CLAUDE.md-inline + AGENTS.md-symlink architecture, multi-agent review, empirical codex verification, and memory cleanup. Inline only — it orchestrates review subagents and runs codex.
Diagnose and resolve Claude Code plugin and skill issues. This skill should be used when plugins are installed but not showing in available skills list, skills are not activating as expected, or when troubleshooting enabledPlugins configuration in settings.json. Triggers include "plugin not working", "skill not showing", "installed but disabled", or "enabledPlugins" issues.
>- Set up and maintain multiple isolated Claude Code CLI profiles so students and power users can run different LLM providers (Kimi K3, Kimi K2.7 highspeed, GLM, DeepSeek, StepFun, Anthropic) in separate terminal windows at the same time. Use this skill whenever the user asks about multi-provider Claude setup, multiple Claude Code windows, switching models or the csk/csd/csg aliases, CLAUDE_CONFIG_DIR, the ~/.claude-profiles directory, or running Kimi/GLM/DeepSeek/StepFun alongside Anthropic. It also owns profile-drift troubleshooting — a third-party profile missing skills/hooks/plugins the default profile has, claude-profiles-doctor reporting a real directory where a symlink belongs, or settings not converging — and per-provider context-window configuration (the [1m] marker or explicit CLAUDE_CODE_MAX_CONTEXT_TOKENS).
Analyze Claude Code and Claude Desktop Code token usage, cost, quota burn, model mix, cache read/write, and 5-hour block consumption using ccusage evidence. Use when the user asks why Claude quota was exhausted, whether a model such as fable/opus/sonnet is unusually expensive, how many tokens were spent today or historically, or needs a human-friendly explanation of local Claude Code CLI/Desktop usage.
Recover actionable context from local `.claude` session artifacts and continue interrupted work without running `claude --resume`. This skill should be used when the user provides a Claude session ID, asks to continue prior work from local history, or wants to inspect `.claude` files before resuming implementation.
Claude Skills — questions
Answers built from the skills we actually parsed.