Find your next open-source project

Search by name, purpose, or keyword — or paste a GitHub repository URL
All skills

oaustegard/claude-skills

A personal collection of Claude skills covering coding, research, media, data and agent management.

  • 116 skills
  • 352 stars

The 5 most installed

Ordered by installs on skills.sh
  • image-to-svg172 installs

    Converts photos, illustrations and line art into SVG vector reproductions.

  • developing-preact168 installs

    Builds Preact web apps with minimal dependencies, including data visualization and accessible widgets.

  • Reads an AI or ML paper and extracts what an enterprise engineer should do about it.

  • asking-questions111 installs

    Asks clarifying questions when a request is ambiguous or a choice could change the outcome.

  • mapping-codebases107 installs

    Deprecated codebase mapping skill, replaced by tree-sitting for exploring repository structure.

The other 111 skills
  • Orients you in an unfamiliar repository, explaining what it contains and how it is arranged.

  • Pulls out the key terms and topics from a document in any of 34 languages.

  • Builds interactive charts as portable Vega-Lite JSON specs from uploaded data.

  • Retrieves clean markdown from pages that normal fetching fails on, using the Jina reader service.

  • exploring-data69 installs

    Profiles and analyzes uploaded datasets, including duplicates, joins, drift and time-series checks.

  • flowing65 installs

    Runs a multi-step procedure as a Python workflow that enforces order, retries and rollback.

  • Helps you decide whether instructions belong in project settings, a skill or a prompt, and structures them.

  • api-credentials56 installs

    Stores and retrieves API keys for providers like Anthropic, Gemini and GitHub when skills need them.

  • Runs a panel of expert perspectives to work through a hard decision or problem.

  • Turns code changes into git patch files you can apply to your local copy.

  • Deprecated alias that forwards to orchestrating-chat-agents so existing script imports keep working.

  • down-skilling52 installs

    Rewrites instructions and examples so a smaller, cheaper model can handle the task reliably.

  • Installs and updates skills from the author's GitHub repository into your skills folder.

  • seeing-images51 installs

    Inspects images closely to describe details, compare regions, read degraded text and reproduce them as SVG.

  • Researches a topic across multiple web sources to build current, well-sourced knowledge.

  • browsing-bluesky49 installs

    Searches posts, reads feeds and lists, and analyzes accounts on Bluesky.

  • Plays music, searches songs and manages your Spotify playlists and library.

  • cloning-project48 installs

    Exports a Claude.ai project's instructions and knowledge files into a bundle you can re-import elsewhere.

  • Writes browser bookmarklets that run JavaScript when clicked from the bookmarks bar.

  • creating-skill48 installs

    Creates or revises an agent skill and packages it as a validated zip.

  • githubbing48 installs

    Installs and authenticates the GitHub CLI in a container, then runs gh commands.

  • Builds searchable progressive-disclosure indexes of GitHub repos for use as project knowledge.

  • iterating47 installs

    Keeps stateful work going across multiple sessions by accumulating context and work logs.

  • Automatically versions every skill file change so you can roll back or compare.

  • Groups Bluesky accounts by topic using keyword extraction from follower and following lists.

  • check-tools46 installs

    Checks that Python, Node, Java, Go, Rust, C/C++, Git and system tools are installed and working.

  • Indexes the Python code inside skills so functions can be imported across skills.

  • Builds production-ready MCP servers with FastMCP, including tool descriptions and packaging.

  • invoking-gemini45 installs

    Calls Google Gemini models for structured output, image generation, speech and multimodal tasks.

  • invoking-github45 installs

    Reads, commits and opens pull requests on GitHub repositories from a chat environment.

  • using-webctl45 installs

    Automates a headless browser to visit pages, scrape content and run Playwright scripts.

  • hello-demo44 installs

    Delivers a static Hello World HTML demo page with a bookmarklet.

  • making-waffles44 installs

    Writes WAFFLES declarations listing what a social media post does not say, to prevent misreading.

  • configuring43 installs

    Loads secrets and configuration from Claude.ai, Claude Code, Codex, Jules and .env files.

  • Splits complex tasks into skill-typed subtasks, routes only the needed context, and runs subagents in parallel.

  • tree-sitting31 installs

    Answers where a symbol is defined, what it spans and who references it, with exact line ranges.

  • Forecasts future values from time series data using the Reverso model on CPU.

  • charting29 installs

    Picks the right Python library and produces publication-quality static charts, plots and diagrams.

  • Resizes, crops, converts, compresses and composites images, including GIFs, montages and metadata inspection.

  • declauding27 installs

    Rewrites drafts so they read like human writing instead of AI-generated prose.

  • processing-video27 installs

    Converts, trims, merges and compresses video and audio with ffmpeg, including GIFs, subtitles and loudness.

  • Sorts your grocery list into store aisle order using photos of aisle signs.

  • Finds every real caller or definition of a Python symbol, excluding same-named false positives.

  • tiling-tree23 installs

    Splits a problem into exhaustive non-overlapping parts and evaluates every option found.

  • featuring21 installs

    Writes a feature inventory explaining what a codebase does from a user's point of view.

  • assessing-impact19 installs

    Reports what would break before you rename, refactor or delete a symbol in a repo.

  • finding-skills19 installs

    Lists, searches and shows installed skills so the agent can load the right one on demand.

  • mapping-webapp19 installs

    Documents a web app's features and behavior by reading its source code, with optional visual checks.

  • Checks patches, locates bugs and compares fixes by tracing execution across files and modules.

  • Rewrites prompts and instructions so the model actually follows them.

  • Builds navigable semantic maps of PDF papers, specs and legal documents with page-anchored claims and symbols.

  • session-memory18 installs

    Keeps a running notes document of decisions, corrections and files touched during long sessions.

  • challenging17 installs

    Runs an adversarial review of a deliverable to catch errors before you ship it.

  • coding-mojo17 installs

    Installs, compiles and runs Mojo code in Claude.ai containers, including benchmarks against Python.

  • llm-as-computer17 installs

    Runs and traces programs on a compiled transformer stack machine.

  • tracking-todos17 installs

    Maintains a structured task list for the current session that survives restarts.

  • container-layer16 installs

    Builds and caches a custom container environment so installs and config survive ephemeral sessions.

  • Finds image regions from text labels like "dog face" to produce bounding boxes automatically.

  • Teaches a person how a codebase works through guided exercises and an HTML learning artifact.

  • semantic-grep16 installs

    Finds text or code by meaning rather than exact keywords, so fuzzy searches still return matches.

  • uploading-files16 installs

    Lets you send files into a Claude Code web session through a temporary GitHub branch.

  • closing-issues14 installs

    Closes a GitHub issue with a synthesis comment, then runs follow-up work in the background.

  • composing-html14 installs

    Builds single-file HTML reports, decks, flowcharts and prototypes from a small JSON spec.

  • Installs and drives Google's Antigravity CLI as a non-interactive sub-agent.

  • opening-prs14 installs

    Creates a GitHub pull request through the API, handling branch, push and mergeable checks without git.

  • bm2512 installs

    Ranks and searches a text corpus by relevance using BM25 scoring.

  • converting-files11 installs

    Converts documents, images, audio and video from one file format to another.

  • Draws images, vector files and animation frames in Python, including shaped text and soft effects.

  • json-render-ui10 installs

    Generates dashboards, admin panels and forms from natural language as constrained JSON.

  • Checks generated music by rendering it, reading spectrograms and pitch, and engraving it as a score.

  • Reads and transcribes what is shown in slides, scanned pages, charts and image-only documents.

  • agent-routing9 installs

    Chooses the model, effort level and structure for each subagent based on measured cost per task.

  • Revises an existing skill through measured, validated edits when there is evidence it underperforms.

  • Extracts headings, sections, code blocks and links from Markdown files with a jq-like query tool.

  • Checks whether a README, guide or spec still matches the code and tests, and reports mismatches.

  • python-lsp8 installs

    Answers Python code questions like go-to-definition, find-references and inferred types using pyright.

  • creating-kb7 installs

    Turns your files into a self-contained, searchable knowledge base bundle any agent can use.

  • Slices photos of many business cards, reads each card, and writes deduplicated contacts to a CSV.

  • creating-video6 installs

    Generates a multi-scene video from a prompt, reviewing each clip and assembling the final cut.

  • parsing-video6 installs

    Samples video frames into timestamped contact sheets so the agent can describe scenes and actions.

  • An older, deprecated version of the GitHub repository index builder; use building-github-index instead.

  • drawing-video5 installs

    Turns a short video into a narrated comic strip with generated panels.

  • Finds tests that only cover a hand-picked subset of a domain and names the missing cases.

  • Chooses how to launch many subagents in parallel and runs the fan-out from one command.

  • Helps you organize symptoms and weigh evidence before a medical appointment.

  • muninn-boot4 installs

    Sets up the Muninn utilities on the container Python path at the start of a new conversation.

  • ocring-pdfs4 installs

    Adds a searchable text layer to scanned PDFs so you can select, search and hand them to other tools.

  • Vectorizes a portrait or photo so the subject keeps more detail than the background.

  • atprotoing3 installs

    Reads Bluesky and ATProto data directly, including likes, replies, backlinks and outages, without the AppView.

  • Classifies, routes or rates text with a calibrated confidence score for each option.

  • Briefs a subagent with just the task while automatically supplying the relevant prior context.

  • Screenshots a Bluesky post with the author's identity blacked out so you can quote it anonymously.

  • gating3 installs

    Builds and audits verification checks that can be shown to fail when results are wrong.

  • Assigns items to a large closed label vocabulary without sending the whole schema to the model.

  • Runs many Claude calls in parallel over the API from claude.ai, with shared cached context and named agent pools.

  • Runs many cheap coding workers in parallel to fix a large backlog of failing tests.

  • Creates short animated films by writing drawing code that renders frames to MP4.

  • Writes a feature-by-feature product description from the code and tests, then checks it against the running product.

  • Finds bugs by testing code against a reference model, then shrinks failures into regression tests.

  • Records console logs, errors, network traffic and DOM changes from a user's browser page for debugging.

  • compiling-wgsl1 installs

    Converts GLSL to WGSL, minifies WebGPU shaders and proves the output unchanged.

  • Builds and maintains a Bluesky curated list and turns its weekly posts into a digest.

  • Improves a program against your scorer by running cheap model swarms within a dollar cap.

  • Finds inputs that crash a parser or shader using the Z3 solver.

  • Builds MP4 videos from a JSON scene file with Kinescript, including narration and captions.

  • Builds MP4 videos from React components with Remotion, including app demo reels inside a phone frame.

  • Lets you draw on a live web page in Chrome and reads back what each mark points at so it can be fixed.

  • Checks each code review finding against the cited code and keeps only the real ones.

  • Reads and writes GitHub repositories through the REST API when cloning fails or repos are private.

  • Breaks a stuck problem frame by committing to one structured creative move before judging ideas.