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. 79 341 files from 1 736 authors, of which 61 700 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 building Android UIs with Jetpack Compose, managing state with remember/mutableStateOf, or implementing declarative UI patterns.
React and Next.js performance optimization guidelines from Vercel Engineering. This skill should be used when writing, reviewing, or refactoring React/Next.js code to ensure optimal performance patterns. Triggers on tasks involving React components, Next.js pages, data fetching, bundle optimization, or performance improvements.
Guardrails for AI agents editing @posthog/warlock, adding or porting rules, proposing new categories, or reviewing Warlock PRs. Load when working in /warlock or on a PR that touches the Warlock source, rules, or docs.
Preview or publish the current Codex coding session as readable Markdown in a configured GitHub repository. Use when the user asks to share, export, publish, or save a Codex session or transcript.
Write and review unsafe Rust `# Safety` documentation and safety comments (e.g., `// safety:` or `/// safety:` in any capitalization) as proof obligations grounded in the Rust Reference, standard library documentation, trusted opted-in dependency contracts, and explicit project invariants.
Active, Socratic teacher for the LangChain Academy Deep Agents course — drives the student through the curriculum with a calibrated interaction density, explain-then-check dialogue at load-bearing moments, and misconception-targeted questioning
Graph database skills for Apache AGE on PostgreSQL. Covers the full lifecycle: deriving an ontology from structured or unstructured data with a human feedback loop, building the graph, and querying it with openCypher, natural language to Cypher, and graph augmented retrieval.
Expert PostgreSQL skills with intelligent routing. Covers both generic PostgreSQL and Azure Database for PostgreSQL.
|- Use this skill when you need to set up validation for AI agent skills in a Dart project for the first time. Adds the linter as a dev_dependency, creates a configuration file, and generates a baseline for legacy repos.
> Validates an in-progress PR or feature branch of skills_lint against known downstream ecosystem consumers. Use when assessing breaking changes across external repositories during PR evaluation, testing migrations against the changelog, or determining necessary backwards compatibility shims.
> Instructions for adding a new validation rule and CLI flag to skills_lint. Use this skill when asked to create a new rule that validates aspects of skills (like frontmatter metadata).
|- Use this skill when you need to validate AI agent skills with skills_lint — running the linter, interpreting failures, fixing violations, and authoring custom rules.
Write and organize unit tests for functions, methods, and classes using `package:test`. Use when creating new logic or fixing bugs to ensure code remains correct and regression-free.
Mandatory checks to run before completing any task that touches md files or dart code in this repository.
Use when implementing any feature or bugfix, before writing implementation code
> How to integrate, update, and configure the skills_lint validation tool within a repository. Make sure to use this skill whenever the user asks to update skills_lint, configure skills validation tests, fix skills linter dependency drifts, verify repository state before editing, optimize lint rules execution, or draft pull request submission commands.
>- De-risk the risky assumptions in a vision by running spikes. Given the revisioner risky-assumptions.yaml, this designs a spike for each open assumption (literature/evidence search, hands-on trial, or a fresh empirical investigation), runs the spikes in parallel isolated workspaces, captures every artifact under .amplifier/revisioner/data/<id>/, and updates each assumption's signed confidence from the evidence gathered. Use when the user wants to validate/invalidate assumptions, "run spikes", gather evidence, or "de-risk" a plan. It never touches the risk (stakes) axis — only confidence.
>- which risky assumptions need the user's attention right now (amend the vision, or unblock a spike), and a drill-down into the evidence behind every verdict. Reads .amplifier/revisioner/ (risky-assumptions.yaml plus each data/<id>/ spike directory), emits a single revision-state.json, and instantiates a bundled React template the user runs with pnpm or npm. Use this whenever someone wants to SEE the state of the "visualize the assumptions", "build a dashboard for this", "is there a UI for the revisioner", "I want to click into the evidence", or after a de-risking pass when a text report is too dense to reason about. Also use it when the user wants to share the current state with someone who will not run the tool themselves.
>- Check in on where a vision stands once its risky assumptions have been found and are being de-risked. Reads the revisioner risky-assumptions.yaml plus the live per-spike status.json files, classifies every assumption (holds / fails / all-clear when the high-risk assumptions hold, ask the user to supply what blocked spikes need (then resume de-risking), or propose a vision update/pivot when evidence has invalidated a high-risk assumption. Use when the user asks "where do we stand", "check in on the vision", "what's blocked", "are we de-risked yet", or after de-risking spikes have run. It reads state and steers; it never sets risk or confidence itself.
>- Surface the risky, unvalidated assumptions hidden inside a vision, product plan, proposal, solution write-up, spec, or design doc, and record any NEW ones in the project's risky-assumptions file. Use this whenever the user has a vision/plan/ proposal artifact and wants to know what could be wrong, what is being taken for granted, what is risky or shaky, or what the author has not fully thought through — or asks to "find/surface/extract risky assumptions", "what are we assuming", "poke holes in this", "pre-mortem this", or to run the ReVisioner assumption pass. Examines the doc from feasibility, viability, and desirability angles (technical, practical, cost, operational, adoption) and appends only assumptions not already captured, without ever changing existing risk levels. This skill only FINDS and records assumptions; de-risking them happens in a separate skill.
Claude Skills — questions
Answers built from the skills we actually parsed.