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. 79 404 files from 1 741 authors, of which 61 763 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 skill should be used for project-level decisions about LLM-powered systems: whether an LLM is the right primitive for the task at hand, the shape of a multi-stage batch or agent pipeline, token and cost estimation, choosing between single-agent and multi-agent at the project level, structured output design for downstream parsing, and structuring agent-assisted iteration. Use this when the unit of work is a whole project or a multi-stage pipeline. Route individual tool design to tool-design and individual skill-loading or context-budget tactics to context-optimization.
This skill should be used when the harness, scaffold, workflow, or optimizer itself is the optimization target: recursive self-improvement (RSI) loops, meta-harnesses, self-improving harnesses that mine their own failures and propose bounded edits, evolutionary or population-based search over agent scaffolds, acceptance gates for self-modifying systems, and agentic context evolution where the mechanism that produces context is versioned and evolved. Route governance of a single autonomous loop (locked surfaces, durable logs, rollback, novelty gates, approval boundaries) to harness-engineering, measurement and quality-gate design to evaluation, judge design to advanced-evaluation, and remote sandbox infrastructure to hosted-agents.
This skill should be used for the tool-interface layer of an agent system specifically: writing tool descriptions agents can route on, designing tool schemas and response formats, naming conventions, actionable error recovery messages, MCP server design, tool-set consolidation, and deciding when to add or remove an individual tool. Use this when the unit of work is a single tool or a set of tools. Route project-shape, pipeline architecture, and task-model-fit decisions to project-development; route deciding whether to introduce sub-agents to multi-agent-patterns.
Guide a new user through their first computer-use agent run (env check, safe browser-only task, then open the trajectory viewer).
Investigate a failing GitHub Actions run or job and create a GitHub issue for the failure.
Linting and formatting for Megatron-LM. Covers running autoformat.sh, tools (ruff, black, isort, pylint, mypy), and code style rules.
How to launch distributed Megatron-LM training jobs on a SLURM cluster. Covers a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules across hardware and parallelism modes, container conventions, monitoring, and per-rank failure diagnosis.
Split a PR into multiple PRs to reduce the number of required CODEOWNERS reviewer groups.
Test system for Megatron-LM. Covers test layout, recipe YAML structure, adding and running unit and functional tests, golden values, marker filters, and CI parity.
Automate React Native Windows integration with upstream React Native nightly versions. Use when: upgrading RNW to a newer React Native nightly, finding target nightly versions, preparing integration PRs, updating package dependencies.
Onboard 1-node GitHub MR functional tests for GB200 from existing mr-scoped 2-node tests.
Migration guide for moving Megatron Core GPTModel checkpoints, model providers, training commands, and layer mappings to HybridModel.
Bump the NVIDIA PyTorch base image (`nvcr.io/nvidia/pytorch:YY.MM-py3`) used by Megatron-LM CI. Covers the two pin sites (GitHub CI in `docker/.ngc_version.dev` and GitLab CI in `.gitlab/stages/01.build.yml`), the post-bump CI loop (re-run functional tests, refresh golden values, mark broken tests), and the gotchas that bit PRs #4611 and #4688.
CI/CD reference for Megatron-LM. Covers CI pipeline structure, PR scope labels, triggering internal GitLab CI (which force-pushes the current branch to a pull-request/BRANCH ref — always dry-run and verify the destination first; never run against shared or protected branches), and CI failure investigation.
Container-based dev environment setup and dependency management for Megatron-LM. Covers acquiring and launching the CI container, uv package management, and updating uv.lock.
Research and draft a response to a GitHub issue or question from an external contributor.
Domain knowledge for the nightly main-to-dev sync workflow. Covers merge strategy, CI architecture, failure investigation, and known issues.
Refresh golden values from a GitHub Actions workflow run (failing-only or all jobs), score the change with average normalized relative differences, and produce a PR-ready summary. Use when the user asks to update goldens for a CI run, refresh golden values from a workflow ID, or generate a golden-value diff summary for a PR description.
Operate FrontendExtension (FE) resources in frontend-forge: create, update, rebuild, inspect package artifacts, download packages, publish, unpublish, delete, and debug package/publish controller behavior. Use this skill whenever the user mentions FE operations, FrontendExtension lifecycle, extension package/download/publish/unpublish, artifact ConfigMaps, package Jobs, publisher Jobs, publish target ConfigMaps or Secrets, rebuild-token, package-state/publish-state labels, or troubleshooting FE status in a Kubernetes cluster.
Operate FrontendIntegration resources and the frontend-forge extension. Use when Codex needs to create a FrontendIntegration from FrontendIntegration YAML, update or patch FI lifecycle state, inspect or troubleshoot FI build output, or create, enable, disable, uninstall, and inspect the frontend-forge extension through its InstallPlan and extension resources.
Generate canonical FrontendIntegration YAML from a simplified single-menu authoring model for frontend-forge.
Use when deploying KubeEye for cluster inspection, creating InspectRule/InspectPlan resources, or retrieving inspection results. Covers InstallPlan-based deployment, OPA/PromQL/FileChange/Sysctl/Systemd/NodeInfo/FileFilter/ServiceConnect/CustomCommand rule types, and report retrieval. Always consult this skill when the user mentions KubeEye, cluster inspection, InspectRule, InspectPlan, or inspection reports.
KubeSphere cluster query Skill (read-only). Use when user requests to view cluster list, cluster status, cluster details, or cluster version info. Do not perform any write operations (create, delete, update). Use KubeSphere Console for management operations.
KubeSphere central controller Skill. Routes to specific Skills based on user requests: multi-cluster management (kubesphere-cluster-management), multi-tenant management (kubesphere-multi-tenant-management), extension management (kubesphere-extension-management). Also provides core architecture, API routing, and API utilities.
Use when configuring ArgoCD in KubeSphere DevOps, including GitOps deployments, application management, SSO setup, or troubleshooting ArgoCD issues
Use when managing credentials in KubeSphere DevOps, including repository credentials, kubeconfig, and API tokens
Use when configuring Jenkins in KubeSphere DevOps, including agent customization, LDAP/OIDC integration, build artifact retrieval, or troubleshooting Jenkins issues
Use when working with KubeSphere DevOps extension, CI/CD pipelines, Jenkins integration, or pipeline troubleshooting
Use when creating, running, or managing CI/CD pipelines in KubeSphere DevOps, including pipeline API operations and run monitoring
Use when operating KubeSphere DevOps as a namespace-scoped tenant with limited permissions, without cluster-admin access, or when accessing DevOps through KubeSphere APIs only
KubeSphere extension management Skill. Use when user requests to install, configure, upgrade, uninstall extensions, or query extension info/troubleshoot issues. Includes extension discovery, dependency management, install configuration, version management.
KubeSphere Fluid management Skill. Use when user asks to install or enable Fluid, check Fluid status, view Fluid pods/logs/CRDs, create or update Dataset, AlluxioRuntime, JuiceFSRuntime, or ThinRuntime, perform DataLoad or cache warming, scale runtime, or troubleshoot Fluid issues in KubeSphere.
KubeSphere Gateway API extension management Skill (Traefik based, uses Kubernetes Gateway API + GatewayProxy CRD gatewayapi.kubesphere.io/v1alpha1). This is the newer Kubernetes Gateway API standard. For the older Ingress API based gateway (ingress-nginx + Gateway CRD gateway.kubesphere.io/v2alpha2), see the kubesphere-gateway skill instead. Covers installation, uninstallation, status checks, GatewayProxy status inspection, and troubleshooting.
KubeSphere Gateway extension management Skill (ingress-nginx based, uses Kubernetes Ingress API + Gateway CRD gateway.kubesphere.io/v2alpha2). For the newer Kubernetes Gateway API (Traefik + GatewayProxy CRD), see the kubesphere-gateway-api skill instead. Covers installation, uninstallation, status checks, gateway status inspection, and troubleshooting (gateway stuck states, Helm failures, pod issues).
KubeSphere multi-tenant management Skill. Use when user requests to create users, workspaces, projects, or assign roles/permissions. Supports user lifecycle management, workspace configuration, project creation, role binding. Do not perform any delete operations, do not create custom roles.
Operate the KubeSphere network extension. Use when Codex needs to install, upgrade, configure, enable, disable, or inspect the `network` extension; manage Calico `IPPool` resources, namespace bindings, migrations, or network isolation flows; or consult the bundled network extension references in this skill.
KubeSphere OpenKruise management Skill. Use when user asks to install or enable OpenKruise, check OpenKruise status, view kruise pods/logs/CRDs, create or update SidecarSet, manage sidecar injection, create or update CloneSet, perform in-place update or batch rollout, uninstall or remove OpenKruise, or troubleshoot Kruise Pod, CRD, and webhook issues in KubeSphere.
KubeSphere ServiceMesh extension management Skill (Istio + Kiali + Jaeger). Use this skill for the ServiceMesh extension Configuration (covers installation, uninstallation, status checks), troubleshooting (covers grayscale release, sidecar injection, topology/metrics, and tracing for Composed Apps Aka Custom Applications).
KubeSphere Volcano job management Skill. Use when user asks to create, list, update, delete Jobs (Volcano Jobs), manage Queues, create PyTorch/TensorFlow/MPI training jobs, or troubleshoot Volcano scheduling issues in KubeSphere. Includes built-in YAML templates, scheduling policy recommendations, and best practices for resource configuration. Handles both KubeSphere API and kubectl operations.
NodeGroup operation Skill for the edgewize nodegroup project. Use this whenever the user wants to query, create, update, delete, bind, unbind, or troubleshoot NodeGroup resources, including node binding, namespace binding, workspace binding, and deployment/config inspection for nodegroup.
KubeSphere OpenPitrix application management Skill. Use when users ask about KubeSphere App Store, OpenPitrix, Helm/YAML application templates, application repositories, app versions, app releases, categories, review states, repository sync, or troubleshooting application installation and upgrade issues.
Use when installing or configuring the OpenSearch extension for KubeSphere, which provides distributed search and analytics engine for storing logs, events, auditing, and notification history
Use when installing or configuring the WizTelemetry Data Pipeline (vector) extension for KubeSphere, which provides data collection, transformation, and routing for observability data including logs, auditing, events, and notifications
Use when working with WizTelemetry Auditing extension for KubeSphere, including installation, configuration, and audit query API
Use when working with WizTelemetry Events extension for KubeSphere, including installation, configuration, and event query API
Use when working with WizTelemetry Logging extension for KubeSphere, including installation, configuration, and log query API
Use when working with WizTelemetry Notification extension for KubeSphere, including installation, configuration, troubleshooting, notification channel setup, alert routing, and silence management.
Use when working with WizTelemetry Ruler extension for KubeSphere, including installation, configuration, alerting rules management
Use when installing or configuring the WizTelemetry Platform Service extension for KubeSphere, which provides the common APIServer backend services for all WizTelemetry observability extensions
Use when working with WizTelemetry Tracing extension for KubeSphere, including installation, configuration, and tracing query API
Always use browser-harness for any web interaction: automation, scraping, testing, or site/app work.
Shape output for a reader with ADHD: lead with the next action, number multi-step work, restate state across turns, suppress tangents, give specific time estimates, make wins visible. Invoke with /i-have-adhd; stays on until "stop adhd mode".
>- Review a GitHub pull request in the googleapis/mcp-toolbox repo against the logic errors and unhandled edge cases, breaking changes, test coverage, docs updates, security (input handling), and new dependencies. Use whenever a maintainer asks you to review, look over, "take a look at", or check whether something is ready to merge in mcp-toolbox, e.g. "review #3703", "can you look at this PR", "is this good to merge", or when they paste an mcp-toolbox PR link. approves, requests changes, comments, labels, or merges on its own.
>- labels (type / priority / product / status), check for duplicates, verify a bug has enough info to act on, and draft a triage comment. Use whenever a maintainer asks you to triage, label, categorize, prioritize, or "look at" an issue (or a batch) in mcp-toolbox, e.g. "triage #3648", "what labels should this get", "is this a dup", triage a batch of issue numbers, or when they paste an mcp-toolbox never edits labels, comments, closes, or assigns on its own.
Deploy, manage, inspect, and troubleshoot Vercel projects from the command line. Use for Vercel deployments, build failures, projects and teams, environment variables, domains and DNS, logs, metrics, Speed Insights, Core Web Vitals, request traces, usage, activity, alerts, firewall rules, cache, cron jobs, deploy hooks, Edge Config, feature flags, integrations, connectors, Blob storage, Container Registry (VCR), microfrontends, rolling releases, custom environments, Sandbox, agent/MCP setup, OAuth apps, preview access, local development, or `vercel api` fallback.
Use for packages/cli changes that affect command UX, prompts, help, output layout, progress, success, warnings, errors, JSON/stdout/stderr contracts, non-interactive/agent behavior, copy, or tests for those surfaces. Do not load for implementation-only refactors with unchanged CLI surface.
>- Turn an exploration (threads, findings, charts) into a single Markdown report — note, blog post, executive summary, KPI dashboard, slide brief, or multi-section analytical report, with embedded charts.
服务端路径安全与文件访问编码规范。在编写文件下载路由、Agent 工具（文件读取/目录列出）、数据连接器/Loader、Workspace 路径操作、沙箱配置时使用。
统一错误处理系统。在添加 API 端点、修改错误处理、添加前端 API 调用、编写错误相关测试时使用。
>- always-available actions (visualize, ask_user, delegate).
Claude Skills — questions
Answers built from the skills we actually parsed.