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. 79 566 files from 1 758 authors, of which 61 913 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.
Identifying vulnerabilities in Node.js dependencies using the npm audit tool to filter for high and critical severities.
Extracting specific security metadata (CVE, CVSS, Fix versions) from structured vulnerability reports like npm audit JSON or OSV.
Use GeoPandas to filter points within a specific polygon, such as finding earthquakes within a specific tectonic plate.
Transform GeoPandas objects to metric projections (like World Equidistant Cylindrical) to calculate distances in meters or kilometers rather than degrees.
Specific knowledge of the PB2002 (Bird, 2002) dataset structure for identifying the Pacific plate.
Calculate the shortest distance between points (earthquakes) and a complex line geometry (plate boundaries).
Strategies for reading and indexing diverse enterprise data formats located in a flat directory or specific path.
Generating a JSON output file with specific requirements for list-based answers and numeric token counts.
How to parse key-value pairs from a text file where keys and question values are stored together.
Methods for calculating or retrieving the number of tokens consumed during an LLM inference task.
Retrieve a fund's AUM, stock holding counts, or compare investment changes between two quarters.
Identify and rank the top fund managers holding a specific stock CUSIP in a given quarter.
Fuzzy search the COVERPAGE data to find the specific accession_number for a hedge fund in a given quarter.
Find the CUSIP (Committee on Uniform Securities Identification Procedures) identifier for a specific company or stock.
Securing Jackson deserialization in Druid to prevent user-supplied JSON keys from overriding system-injected security configurations.
Systematic identification of all Apache Druid components utilizing JavaScript to ensure comprehensive patching of CVE-2021-25646.
Creating, applying, and verifying patches for Apache Druid using Maven with specific flags to bypass non-essential checks.
Processes JSON data from GitHub Search results to compute specific repository health metrics including average merge time, top contributors, and bug resolution counts.
Fetch data from GitHub's Search API using the GitHub CLI (gh) while correctly handling pagination and merging results. Use this when you need to retrieve more than 100 items or ensure full dataset coverage for a specific period.
Extracts text and mathematical descriptions from a PDF file to identify specific algorithm parameters and loss functions. This is used to ensure the implementation matches the theoretical definition in the paper.
Implements the SimPO (Simple Preference Optimization) loss function with length normalization and reward margin as specified in the research paper.
Executes the provided unit test to verify the SimPO loss implementation and saves the output to a specific NPZ file for evaluation.
Aligns the local environment with the repository's requirements by checking for dependency files, installing build-essential tools for complex packages like DeepSpeed, and logging environment metadata.
Manually processes non-standard conditional markers like `{{IF_VARIABLE}}` and `{{END_IF_VARIABLE}}` in a Word document. It identifies blocks of text to either retain (removing markers) or delete (removing the block) based on data.
Replaces placeholders in a Word document across all sections, including headers, footers, body paragraphs, and deeply nested tables. It uses paragraph-level replacement to ensure placeholders split across multiple runs are correctly identified.
Discovers PDF, DOCX, and PPTX files in a directory (handling hidden characters and extensions robustly), extracts their text, determines their subject, and moves them into organized folders.
Categorizes a document into one of five subjects (LLM, trapped_ion_and_qc, black_hole, DNA, music_history) based on an expanded technical keyword dictionary. It prioritizes the beginning of the text (titles and abstracts) where subject density is highest.
Extracts text from .docx and .pptx files using `pandoc`. This tool is preferred for its robustness in handling various document schemas and converting them into plain text for analysis.
Extracts text from PDF files using the `pdftotext` command-line utility with the `-layout` flag to preserve multi-column formatting, which is essential for accurately parsing scientific papers.
Define idiomatic Scala 2.13 type hierarchies and domain models, replacing Python's class structures and Enums with traits and case classes.
Implement a fluent Builder pattern in Scala to provide a clean API for object configuration and instantiation.
Implement idiomatic Scala logic for parsing temporal/numeric data and handling batch processing using functional transformations and error types.
Parse the calendar PDF to identify existing appointments, flexible blue slots, and the 15-minute grid coordinate mapping.
Extract meeting duration, constraints, and metadata from the input JSON file containing email requests.
Generate meeting confirmation text files based on a specific template and save the results in a JSON log.
Find the earliest available meeting slots for multiple requests, treating blue blocks as available and updating the schedule state after each assignment.
Creates the core D3.js logic in visualization.js. This includes data loading, custom market cap formatting (supporting "T" for Trillions), force simulation for sector clustering, bubble generation, and bidirectional table highlighting.
Creates the HTML and CSS files for the stock visualization app. The HTML provides the structure for the side-by-side layout (bubble chart and table), while the CSS handles styling, scrolling, and highlighting logic.
Initializes the directory structure and populates the environment with necessary data and libraries for the D3.js visualization. This skill should be executed before generating any code files to ensure the target paths exist and the D3 library is available locally.
Manage and calibrate General Lake Model (GLM) configuration files (glm3.nml) within specified parameter constraints and physical ranges.
Calculate specific RMSE metrics by merging simulation and observation data using exact datetime and rounded-depth matching.
Extract and transform water temperature data from GLM NetCDF files, handling dynamic layering, masked arrays, and temporal alignment.
Logic for planning itineraries that rely on ground transportation (driving) rather than flights, ensuring travel feasibility and daily activity engagement.
Guidelines for selecting database-compliant restaurants, accommodations, and attractions based on specific user constraints like budget, pets, and cuisine.
A final verification step to ensure the JSON output meets all structural and constraint requirements before delivery.
Converts all extracted keyframe images in a directory from RGB to grayscale and overwrites the original files.
Counts occurrences of specific objects (coins, enemies, turtles) in a frame using a template matching script with optimized parameters.
Converts a video file into a series of keyframe images using FFmpeg with specific frame selection filters.
Aggregates counting data for all frames and objects into a final CSV file formatted as required.
Use standard Excel functions to calculate basic statistical measures across a range of values.
Convert decimal percentages to whole-number representations and round to a specific decimal place.
Perform a two-dimensional lookup in Excel to retrieve data based on both a row criteria (e.g., series code) and a column criteria (e.g., year).
Calculate a weighted mean using the SUMPRODUCT and SUM functions.
Defines the official, sanctioned Anthropic brand palette and HEX mapping for all technical documentation and hardware visualization assets.
Specifies the sanctioned typefaces for Anthropic internal engineering handbooks and technical posters.
Provides the workflow for creating minimalist, low-saturation technical illustrations using the Nova device as a template.
Handles the final writing of the poem into the specified file path, enforcing strict formatting constraints.
Composes an eight-line seven-character regulated verse (Qiyan Lüshi) focusing on historical or emotional themes, ensuring exact line counts and structural integrity.
Applies the specific tonal patterns and rhyme requirements for a seven-character regulated verse to ensure classic prosody.
Use this skill to map specific natural language case details to the legal requirements of the SC-100 form.
Claude Skills — questions
Answers built from the skills we actually parsed.