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. 79 600 files from 1 763 authors, of which 61 947 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.
| This applies to ANY frontend work, not just "design" tasks. Even simple apps benefit from basic design principles.
| Remove signs of AI-generated writing from text to make it sound more natural and human-written. blog posts, PRDs, or any dedicated writing content. Based on Wikipedia's comprehensive promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, negative parallelisms, and excessive conjunctive phrases.
Background knowledge for droid-control workflows -- not invoked directly. Tuistory driver mechanics for terminal TUI automation via virtual PTY.
| Automate browser interactions for web testing, form filling, screenshots, and data extraction. This skill uses agent-browser for comprehensive browser automation.
| This is a meta-skill for self-improvement and continuous learning.
| architecture diagrams, patterns, textures, photo edits, restorations Covers nanobanana CLI for image generation and Slidev for presentations.
| Generate comprehensive codebase documentation for a repository. Uploads the wiki to view in the Factory app.
Analyze code changes for security vulnerabilities using LLM reasoning and threat model patterns. Use for PR reviews, pre-commit checks, or branch comparisons.
Ban `as` type assertions in a package via the `@typescript-eslint/consistent-type-assertions` lint rule, replacing them with compiler-verified type-safe alternatives. Use when enabling the assertion ban in a new package or fixing violations in an existing one.
Scan code changes for security vulnerabilities using STRIDE threat modeling, validate findings for exploitability, and output structured results for downstream patch generation. Supports PR review, scheduled scans, and full repository audits.
Validate security findings from commit-security-scan by assessing exploitability, filtering false positives, and generating proof-of-concept exploits. Use after running commit-security-scan to confirm vulnerabilities.
Generate a STRIDE-based security threat model for a repository. Use when setting up security monitoring, after architecture changes, or for security audits.
>- Enforce the no-useEffect rule when writing or reviewing React code. ACTIVATE when writing React components, refactoring existing useEffect calls, reviewing PRs with useEffect, or when an agent adds useEffect "just in case." Provides the five replacement patterns and the useMountEffect escape hatch.
| (extract to new file), dead barrel re-exports (remove from index.ts), and internally-only-used exports (un-export). Use when `npm run knip` reports unused exports.
Build and debug SceneKit scenes where one 3D object (a product, badge, coin, wheel) performs on a transparent stage inside a SwiftUI app - studio lighting, real shadows, baked keyframe choreography, hand-rolled physics, gestures and haptics, multi-scene sequencing. Use when working with SCNView or SceneView in SwiftUI, UIViewRepresentable 3D scenes, product or hero-object animation, roll or spin entrances, a first-frame hitch when a scene appears, shadows missing or wrong, metal rendering black, choreographed 3D motion that must stay interruptible, a continuous vapor stream (vent air, steam, mist) drawn as a shader-driven sheet, reproducing a real object's motion from photos or video, or keyframed motion that stutters at its own keyframes. Not for RealityKit, ARKit, visionOS, or full game worlds.
Use when text must be easy for its intended reader to find, understand, and act on — reports, emails, documentation, UI text, announcements, explanations, AI output that reads as dense or bureaucratic. Applies ISO 24495-1 plain language principles with dedicated technique layers for English and Traditional Chinese. Triggers: plain language rewrite, apply ISO 24495, make this clearer for readers, de-jargon this, 淺白改寫, 白話改寫, 去公文腔, 讓讀者一次看懂. Not for creative or literary writing.
> 用户想给 DeepSeek Harness 找插件时使用：「有没有插件能……」「帮我装个 XX」 「生态里有什么好玩的」。从全 GitHub 的 dsh-plugin topic 发现跨个人与组织的 公开仓库，筛选候选，等用户拍板，先汇报这个插件要什么权限，再从仓库声明判断 安装方式并验证挂载。只负责找和装；开发新插件转 make-dsh-plugin。
Diagnose a brand, website, or page for evidence-backed GEO gaps and opportunities from user-supplied URLs, HTML, or evidence. Use for brand diagnosis, website or page audits, GEO gap analysis, and 品牌诊断、网站诊断、页面诊断. Excludes live AI-platform recall, ranking, and citation-share measurement.
Measure GEO visibility from an approved, file-backed engine observation bundle. Use for AI answer mention rate, source inclusion, citation share, query-panel coverage, GEO monitoring, 监测 AI 可见度, 衡量 GEO 效果, and offline baseline comparison. Exclude live scraping, platform login, automated collection, and unsupported causal claims.
Discover evidence-aware GEO questions, query rewrites, intent clusters, and prioritized content opportunities from a structured brief. Use for AI search intent mining, question or keyword expansion, query research, FAQ discovery, 拓词, and GEO topic discovery in Chinese or English.
Create evidence-lined GEO titles, explainers, neutral comparisons, method-backed rankings, page blueprints, refinements, and article-friendly Markdown. Use for title generation, 科普/解释、对比、榜单、页面蓝图、内容优化, and article-friendly rewriting. Excludes unsupported factual claims, winner declarations without like-for-like evidence, network research, publishing, and live ranking measurement.
Turn one natural-language request into an evidence-bounded SEO work plan across technical SEO, crawling and indexing, Search Console incidents, keyword-to-page mapping, migrations, experiments, international or ecommerce SEO, and AI-search foundations. Use for 一句话SEO、技术SEO、自然搜索、收录诊断、关键词页面映射 and website SEO. Exclude paid search, ASO, ranking guarantees, link spam, live mutation without authorization, and unsupported metrics.
Build an offline GEO optimization plan from explicit goals, diagnosis actions, approved evidence IDs, and a measured baseline. Use for GEO strategy, roadmap, experiment planning, intervention candidates, 策略, 路线图, and 优化实验. Exclude autonomous publication, fabricated outcome evidence, and memory promotion without positive external measurement.
Route GEO and generative engine optimization requests to an available GEOHub capability. Use for broad GEO requests, GEOHub workflow selection, capability checks, or requests spanning discovery, brand/site/page diagnosis, content, strategy, knowledge, publishing, and measurement.
Build and query an evidence-lined GEO knowledge graph from approved source bundles. Use for entity normalization, relation lineage, conflicting fact preservation, incremental source-hash updates, local/global knowledge queries, 知识图谱, 知识库, and 知识治理. Exclude unsourced facts, hidden conflict resolution, autonomous crawling, and external database mutation.
Turn 2+ user reference images into a high-fidelity animated Codex companion using PocketMen's own local stack. Prefer the open-weight Neural Local Studio (FLUX.2 klein 4B; optional Qwen-Image-Edit-2511 Identity-Max) on compatible hardware; otherwise fall back to the deterministic local renderer. Never require hatch-pet or OPENAI_API_KEY for the normal workflow.
> Generate North American road-trip itineraries as a map-first, offline-friendly single-file HTML page. Plans around daily driving segments, overnight stops, fuel/EV-charging, national-park reservations (Recreation.gov / NPS), seasonal road closures, and timezone/border crossings — for executable, decision-ready the whole route, or hand it an existing route and it verifies, fills gaps, and produces the page.
Generate transcriptome-wide virtual spatial transcriptomics from H&E histology with DeepSpot-M. Use when you need spatial gene expression in log1p-CPM for 224x224 tiles at about 20x, want to query protein-coding genes by symbol instead of a fixed panel, or want to run prediction across a whole slide after tiling with histolab.
Query and download public cancer imaging data from NCI Imaging Data Commons. Invoke for any question about IDC collections, cancer imaging datasets, DICOM data access, radiology (CT, MR, PET) or pathology AI training sets, metadata queries, visualization, or license checks — even when the user doesn't explicitly mention "IDC". No authentication required.
Design custom laboratory hardware as parametric build123d models and export fabrication-ready STEP, STL, and DXF files - microfluidic chips and molds, optomechanical mounts and breadboard adapters, cuvette and microplate holders, tube racks, animal-behavior rigs, and 3D-printed instrument fixtures. Use when a research task needs a physical part that must mate with standardized labware, an optical table, a cage system, or a printer, CNC, or laser process.
Queries the NCATS Translator ARAX production API for bounded, typed, provenance-rich one-hop and endpoint-pinned two-hop biomedical knowledge-graph relationships. Use for Biolink-constrained RTX-KG2 lookup, explicit selected-provider ARAX federation, separate entity normalization, qualifier-aware graph traversal, and inspection of TRAPI edge bindings, publications, and knowledge-source provenance. Do not use for inference, ranking, open-ended pathfinding, clinical guidance, or sensitive queries.
GPU-accelerates scientific Python on NVIDIA hardware and verifies that the result is correct and faster. Use for CUDA/GPU optimization; CPU-bound NumPy, SciPy, pandas, scikit-learn, NetworkX, scikit-image, vector-search, image-processing, graph, simulation, or file-I/O workloads; CuPy, cuDF, cuML, cuGraph, cuVS, cuCIM, KvikIO, Warp, Newton, Numba-CUDA, or RAFT questions; and profiling, memory-transfer, kernel, or multi-GPU bottlenecks. Also use when large data-parallel Python code is slow and GPU acceleration is a plausible option, even if the user does not name CUDA.
Build with and use Pi, the minimal terminal coding harness. Use for installing Pi, configuring providers/models/settings/environment variables, creating Pi skills/extensions/packages/themes/prompt templates, embedding Pi through the SDK, integrating over RPC or JSON event streams, parsing sessions, running local models through the llama.cpp router, developing custom Pi providers and TUI components, or using ecosystem packages such as pi-subagents (delegation/orchestration), pi-mcp-adapter (MCP servers), pi-interview (interactive forms), and pi-web-access (web search, fetching, video understanding).
Multivariate severity assessment and humane endpoint prediction for laboratory animal studies using the RELSA (RELative Severity Assessment) score and ARIMA-based foRcast forecasting. Use when combining welfare readouts — body weight or weight loss, body temperature, clinical or nesting scores, biomarkers, activity, heart rate, burrowing, wheel running — into one severity score per animal per day, when asking which animals are at risk of reaching a humane endpoint or when one will be reached, when defining attention/danger zones or thresholds on a severity scale by kernel density estimation, or when reporting severity for a 3Rs, refinement, animal-welfare, or EU Directive 2010/63/EU severity-assessment context. Covers directionality ("turned" variables), baseline normalization, reference sets, RELSA weights, ARIMA prediction intervals, and RMSE/PICP/MPIW evaluation.
Rowan is a cloud-native molecular modeling and medicinal-chemistry workflow platform with a Python API. Use for pKa and macropKa prediction, conformer and tautomer ensembles, docking and analogue docking, protein-ligand cofolding, MSA generation, molecular dynamics, permeability, descriptor workflows, and related small-molecule or protein modeling tasks. Ideal for programmatic batch screening, multi-step chemistry pipelines, and workflows that would otherwise require maintaining local HPC/GPU infrastructure.
Production-ready reinforcement learning algorithms (PPO, SAC, DQN, TD3, DDPG, A2C) with scikit-learn-like API. Use for standard RL experiments, quick prototyping, and well-documented algorithm implementations. Best for single-agent RL with Gymnasium environments. For high-performance parallel training, multi-agent systems, or custom vectorized environments, use pufferlib instead.
Use when working with Outpost Bio's open microbiome foundation models - the Waypoint checkpoints (Waypoint-6m, Waypoint-45m, Waypoint-170m), the Atlas pretraining corpus, the Compass eight-task benchmark, or the `waypoint` CLI from the `waypoint-bio` package. Covers embedding microbiome samples, fine-tuning on taxonomic abundance data, benchmarking a checkpoint on Compass, pretraining a GPT-2 model on taxonomic abundance profiles, and converting MetaPhlAn, Kraken2, QIIME 2, or MGnify abundance tables into waypoint format.
Run deadline-bound or scope-sensitive agent tasks with explicit done conditions, protected verification time, evidence gates, and safe-stop rules. Use for rapid implementation, debugging, integration, delegated work, feasibility gates, or tasks where scope drift, false completion, repeated failure, or deadline overruns are material risks.
Generate, revise, and audit source-backed quantitative manuscript figures, and visually audit exported schematic/conceptual figures and graphical abstracts without redrawing them in v0.1. Use for Elsevier-style manuscript figure generation, matplotlib or ggplot2 result plots, line charts, heatmaps, bar charts, scatter or Pareto plots, artwork export checks, pre-submission figure QA, mechanism/workflow diagram audits, graphical-abstract audits, or equivalent Chinese-language requests for paper figures and submission checks.
Create project skeleton. Pick stack, create files, install dependencies. AI decides everything.
When multiple solutions exist, pick the best one. Explain why in one sentence.
Celebrate real milestones only. One line, one emoji. Credit belongs to user, not AI.
Classify request as small/medium/large. Adjust workflow depth accordingly.
Build user-facing interface. Clean, functional, mobile-friendly by default.
Silent quality check after every feature. Fix issues before telling user. Never claim tests passed without running them.
Build one feature at a time. Complete each fully before moving to next. Auto-triggers quality check.
Before coding, determine what to ask. Max 2 yes/no questions. Never ask technical questions.
I-Lang compression engine. All internal planning uses I-Lang v5.0 syntax. User never sees compressed output.
End of session summary. What got done, what got fixed, what comes next, progress delta.
Translate technical decisions into human language. Explain in cost, speed, stability.
Deploy to Cloudflare Workers. Free tier handles 100k requests/day. Global edge network.
Deploy to VPS. Code is already on the server. Start the service, configure nginx, verify accessible.
Explain all costs in human terms. Always compare with real-world equivalents. Recommend cheapest that works.
Choose deployment target based on project type. Static sites to CF Pages, APIs to VPS, serverless to Workers.
Help user buy a domain, configure DNS, set up SSL. Guide every click.
Help complete beginners set up their development environment. Detect Mac or other. Guide VPS purchase and SSH setup step by step.
At milestones, compare achievement vs human programmer time and cost. Keep it realistic.
Auto-fix bugs. Observe symptom, find root cause, apply minimal fix, verify, explain in human terms.
Transfer files between local and server. Guide user through SCP or upload methods.
Step 1 of debugging: observe the symptom carefully before jumping to conclusions.
Claude Skills — questions
Answers built from the skills we actually parsed.