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.
Finding the Pareto frontier for multi-objective optimization using dominance conditions.
An advanced guide on generating a fully-featured CSV vulnerability report, featuring explicit filtering, missing data fallbacks, and strict header schemas.
Advanced robust CVSS extraction logic that correctly falls back across priority sources and handles missing values.
A complete guide to executing Trivy in offline mode via Python `subprocess` to scan dependency files for vulnerabilities.
Advanced handling of GeoJSON files in GeoPandas, emphasizing correct geometry formats and time processing to ISO 8601 standard.
Advanced guide on GeoPandas projections for accurate metric distance calculations without distortion using point-centric Azimuthal Equidistant CRS.
Improved guide for performing spatial joins with GeoPandas, specifically handling antimeridian crossing polygons like the Pacific Plate.
Deep parsing and structural querying of nested JSON documents, resolving implicit relations across multi-modal corporate datasets like Slack messages and PRs.
How to correctly and robustly search for a fund manager's accession number in SEC 13F COVERPAGE.tsv.
How to calculate a fund's actual AUM and total unique stock count for modern SEC 13F data.
How to accurately identify stocks that "received increased investment" by checking both share count and dollar value change.
How to find top fund managers heavily invested in a particular stock across all data quality issues.
Security considerations for Jackson JSON deserialization in Java, focusing on @JacksonInject bypass vulnerabilities via empty property names ("").
Provides a robust python module for querying GitHub's search API using urllib with automatic pagination.
A refined skill to implement SimPO (Simple Preference Optimization) loss in PyTorch, precisely matching the official paper implementation without a reference model.
Advanced Word document manipulation using raw XML regex replacements to perfectly preserve formatting across split text runs.
Robustly extracts text from PDF, DOCX, and PPTX files in Python using PyPDF2, python-docx, and python-pptx, suitable for document classification workflows.
Classifies unstructured text into specific categories based on keyword presence, falling back to a default category when no matches are found.
Translating Python @dataclass and Enum to Scala case class and Enumeration/Sealed Traits safely
Translating Python generics (TypeVar covariant/contravariant) to Scala generics (+T, -T) with examples
Parsing and handling JSON safely in Scala using io.circe
Translating Python Union types and @overload to Scala method overloads, type classes, and pattern matching
A refined output generation skill that precisely templates text files and correctly constructs the requested JSON log.
An improved skill to programmatically parse PDF calendars, mapping drawing coordinates to precise time intervals using PyMuPDF.
An improved scheduling algorithm to find the earliest compatible free slots under time constraints.
Guide to building a polished D3.js (v6) force-directed bubble chart and a synchronized data table, including integrating secondary data for sparklines.
Instructions for modifying GLM parameters using regex and systematically optimizing them within physical limits.
Instructions for safely parsing GLM NetCDF output with netCDF4 and correctly executing an exact datetime + rounded-depth merge.
How to load and filter CSV travel data in Python
How to serialize the final travel itinerary to a JSON file.
Extract key frames (I-frames) from a video file into a sequence of images using FFmpeg.
Convert a batch of images to grayscale in-place using ImageMagick's mogrify command.
Count the number of object occurrences in an image using OpenCV template matching in Python.
How to perform a two-dimensional lookup in Excel using either INDEX/MATCH or VLOOKUP/MATCH.
Calculating and rounding summary statistics like percentiles in Excel.
Understanding scaling and execution of weighted means using SUMPRODUCT.
Creates a minimalist technical exploded-view poster of the Nova edge device for Anthropic's internal engineering handbook. Displays at least 5 hardware layers (Casing, Thermal Unit, PCB, Battery, Interface) with official Anthropic brand colors, typography, and design standards. Outputs a PNG poster and a JSON file with applied brand color hex values and heading font name.
Create a definitive lookup table mapping each character in your poem to its Mandarin pinyin, tone number (1–4), and 平仄 classification (平=level/rising tones 1–2; 仄=falling/entering tones 3–4). Use only authoritative sources like Xinhua Dictionary. This table serves as your verification anchor before composition and prevents tone misclassification errors.
Fill in your poem position-by-position, starting with the 4 rhyme characters (lines 2, 4, 6, 8). For each remaining position, select a character, verify its pinyin and tone in your reference table, confirm it matches the required 平/仄 classification, then move to the next position. Never move to a new line until all 7 positions of the current line pass verification.
Explicitly write out the required tonal pattern for all 8 lines of your chosen regulated verse form (specify which of the 4 canonical forms: 仄起平收 or 平起平收, etc.). Display the target pattern as a visual grid showing positions 1–7 for each line, then use this as your line-by-line verification checklist during composition.
Check that all 4 rhyme characters (lines 2, 4, 6, 8) belong to the same modern Mandarin rhyme family by comparing their finals and tone patterns. Use Xinhua Dictionary or online Mandarin phonetic tool to confirm final identity (e.g., -ing, -ang, -ong) and ensure all 4 characters have identical tone classification (all 平, or all 仄) in the rhyme position.
Use this skill first to discover the actual field names in a PDF form. Extracts and prints all form field names from a PDF file to ensure accurate field mapping before filling the form.
Use this skill to fill the California Small Claims Court form (SC-100) with plaintiff and defendant information, case details, and amounts. Requires actual field names from the PDF — run Extract PDF Form Field Names skill first.
Implement DBSCAN clustering with a custom weighted Euclidean distance metric controlled by shape_weight parameter. Use this skill to cluster citizen science point annotations on Mars cloud images.
Match clustered centroids to expert annotations using greedy nearest-neighbor matching with distance constraints. Compute F1 scores and delta metrics for clustering quality assessment.
Load citizen science and expert annotation datasets, match them by image using file_rad column, handle missing data, and prepare data for clustering evaluation.
Format and write Pareto frontier results to CSV file with proper rounding and column ordering.
Execute grid search over DBSCAN hyperparameters, evaluate each combination across all images, filter by F1 threshold, and identify Pareto-optimal solutions balancing F1 score and delta metric.
Ensure vulnerability records are deduplicated and ordered according to ground-truth requirements.
Extract CVSS score from vulnerability data using the correct priority order (NVD → GHSA → RedHat) and handle nested JSON structure properly. Use case-insensitive field access and correct JSON path navigation.
Filter vulnerability records to include only HIGH and CRITICAL severity levels with case-sensitive exact matching.
Write CSV output with correct field quoting and escaping for special characters, using standard CSV formatting.
Properly handle null, missing, and empty values in vulnerability records with appropriate fallback values.
Correctly identify and extract Trivy vulnerability JSON field names and map them to the required CSV output columns in exact order.
Combine all Pacific plate boundary geometries using `.unary_union` and calculate the distance from each earthquake to this combined boundary geometry. Returns the minimum distance in kilometers for each earthquake.
Orchestrate the full analysis workflow: load data, identify Pacific plate, filter earthquakes and boundaries, project to EPSG:4087, calculate distances, find the furthest earthquake, and save results. Use this as the main execution skill.
Use the `.within()` geometric method to identify earthquakes that are contained inside the Pacific plate polygon. Use this skill to ensure only earthquakes actually inside the Pacific plate are analyzed.
Filter the boundary dataset to include only boundaries relevant to the Pacific plate (where PlateA or PlateB equals the Pacific plate identifier). Use this skill to exclude irrelevant boundaries before distance calculations.
Identify the earthquake with the maximum distance to the Pacific plate boundary from the earthquakes within the Pacific plate. Use this skill to locate the target earthquake for final output.
Examine the plates dataset to find the exact identifier used for the Pacific plate (e.g., "PA", "Pacific", etc.). Use this skill early in the analysis to ensure consistent filtering across boundaries and plates.
Claude Skills — questions
Answers built from the skills we actually parsed.