Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Define knowledge bundles with Google's Open Knowledge Format (OKF) v0.1. Use when the user mentions OKF, Open Knowledge Format, Google's knowledge format, LLM wiki bundles, agent knowledge packs, creating OKF bundles, validating OKF documents, or converting knowledge into the OKF standard. Do not use this skill for unrelated requests; route to the nearest named specialist.
Audits and optimizes a repository's Claude Code configuration so path-scoped rules load lazily and CLAUDE.md stays a lightweight always-loaded index. First maps the real project tree (domains, monorepo package roots, extensions, generated/frozen areas) so every path glob is derived from and validated against actual files — stale globs that match zero files and over-broad ones are caught. Three areas: (1) migrate .claude/rules/**/*.md frontmatter to the native lazy-load `paths:` YAML-array form (away from eager `globs:` or red-flagged CSV `paths:`) with accurate, tree-validated patterns; (2) slim CLAUDE.md files to an index + rule map and delete redundant @import-only shells; (3) audit/install the PreToolUse(Write) hook that works around the read-not-create caveat. Audits first, shows a report, asks for confirmation, then applies. Pass --audit-only to skip applying.
Multi-agent coordination daemon for coding agents. Use for deterministic port claims, session tracking, salvage, file claims, notes, pub/sub, tuple-space coordination, background fleets, and debugging multi-agent failures across Claude Code, Codex, Gemini CLI, Cursor, or Windsurf. NOT for production deploy orchestration, Docker or Kubernetes networking, or cloud service discovery.
Systematic problem-solving techniques for stuck-ness. Techniques: simplification cascade (complexity spirals), collision-zone thinking (innovation blocks), meta-pattern recognition (recurring issues), inversion exercise (assumption constraints), scale game (uncertainty). Actions: simplify, analyze, recognize patterns, invert assumptions, scale thinking. Keywords: problem solving, complexity spiral, innovation block, stuck, simplification, meta-pattern, assumption inversion, scale uncertainty, breakthrough thinking, root cause, systematic analysis, Microsoft Amplifier, debugging approach, creative solution. Use when: complexity spiraling, hitting innovation blocks, seeing recurring patterns, constrained by assumptions, uncertain about scale, generally stuck on problems.
Build type-safe AI agents and graph-based workflows with PydanticAI and PydanticGraph. Agent creation, function tools, capabilities, dependency injection, structured output, streaming, multi-agent patterns, testing, evals, and graph state machines. Use whenever you are building agents, tool-using LLM workflows, or graph-based state machines in Python. Do not use this skill for unrelated requests; route to the nearest named specialist.
Guide wanderers to the right animal for their journey. Perch, tilt your head, chatter about the forest, present the options, and warble the recommendation. Use when helping users choose which skill to use, discovering capabilities, or navigating the ecosystem.
Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Claude ('No, that's wrong...', 'Actually...'), (3) User requests a capability that doesn't exist, (4) An external API or tool fails, (5) Claude realizes its knowledge is outdated or incorrect, (6) A better approach is discovered for a recurring task. Also review learnings before major tasks.
Scaffold the Mimas agent instruction file tree for any repository — AGENTS.md at root, subdomain CONTEXT.md files, and the full agents-docs/ hierarchy (a sibling of any existing docs/, kept separate so human-maintained project docs stay untouched). Every file is tailored to the repo's actual tech stack, git platform, and conventions. Use this skill whenever someone wants to set up agent instructions, onboard a repo for AI-assisted development, add AGENTS.md / CONTEXT.md files, create engineering docs for agents, or mentions "set up agentic repository" or "mimas template". Even if they just say "set up this repo for agents" or "add agent docs", this is the skill to use.
Run `specify init` in the current or target directory to bootstrap a Spec Kit project (pull .specify/ and slash commands); supports multiple AI agents and --script sh/ps. Use when the user says "initialize Spec Kit project", "specify init", or "set up Spec Kit in this repo".
Install the Specify CLI on the host machine (uv tool install or uvx one-time); supports multiple OS, persistent or one-time install, and corporate or restricted-network environments. Use when the user says "install Spec Kit", "install Specify CLI", or "specify command not found".
Prevents premature execution on ambiguous requests. Analyzes request clarity using 5W1H decomposition, surfaces hidden assumptions, and generates structured clarifying questions before work begins. Use at the start of any non-trivial task, or when a request could be interpreted multiple ways. Triggers on "뭘 원하는건지", "요구사항 정리", "clarify", "what exactly", "scope", "requirements", "정확히 뭘", "before we start".
Load PROACTIVELY when decomposing a user request into parallel agent work. Use when user says "build this", "implement this feature", or any request requiring multiple agents working concurrently. Guides task decomposition into parallelizable units, agent assignment with skill matching, dependency graph construction, and result aggregation. The runtime engine handles WRFC chain coordination automatically via <gv> directives.
Help users install, configure, operate, and migrate OpenClaw (gateway, channels, nodes, plugins, model/auth setup). Use when answering OpenClaw setup/debug questions or model migration questions; use the local docs snapshot bundled with this skill as the source of truth. Triggers: openclaw, clawdbot, clawd, clawdhub, gateway, onboard, channels login, whatsapp, telegram, discord, mattermost, pairing, nodes, sandboxing, tailscale, provider migration, Claude to GPT, model swap.
Integrate OpenAI Agents SDK with You.com MCP server - Hosted and Streamable HTTP support for Python and TypeScript. - MANDATORY TRIGGERS: OpenAI Agents SDK, OpenAI agents, openai-agents, @openai/agents, integrating OpenAI with MCP - Use when: developer mentions OpenAI Agents SDK, needs MCP integration with OpenAI agents
Load this skill when the user wants to run an online experiment (live-traffic A/B test) on an LLM application instrumented with Agent Observability: compare two versions of a prompt, model, or behavior using a Datadog feature flag, score metrics from evaluations, and cost and token metrics. Covers creating the flag, wiring the app, creating the experiment, attaching metrics, setting the traffic split, and starting it. Triggers: online experiment, A/B test an LLM app, test a prompt change on live traffic, experiment with feature flag and evaluation score
Build AI phone agents with AgentPhone API. Use when the user wants to make phone calls, send/receive SMS, manage phone numbers, create voice agents, set up webhooks, or check usage — anything related to telephony, phone numbers, or voice AI.
Create, audit, or trim AGENTS.md and its harness mirrors (CLAUDE.md, GEMINI.md, .github/copilot-instructions.md, .cursor/rules) so build/test commands stay accurate and the file stays under each harness's size cap. Use when the user says "create AGENTS.md", "audit AGENTS.md", "trim AGENTS.md", "AGENTS.md is stale", "CLAUDE.md drift", "agent instructions hygiene", or invokes /agents-md. Covers nested AGENTS.md for monorepo packages, nearest-file precedence, and the bundled scripts/agents_md_check.py measurement script. Do NOT use for `.github/copilot-instructions.md` authoring or Copilot code-review configuration — that is /github-copilot-repo-instructions. Do NOT use for personal, per-user, or machine-global instruction files (`~/.claude/CLAUDE.md`, personal Copilot instructions) — those are out of scope for this repository-scoped skill.
Build conversational multi-agent systems with Microsoft AutoGen. AssistantAgent, UserProxyAgent, GroupChat, code execution, nested chats, cancellation tokens, tool integration, and MCP support. Use when building conversation-driven multi-agent systems or comparing agent frameworks. Do not use this skill for unrelated requests; route to the nearest named specialist.
This skill helps an LLM generate correct AxGEPA optimization code using @ax-llm/ax. Use when the user asks about AxGEPA, GEPA, Pareto optimization, multi-objective prompt tuning, reflective prompt evolution, validationExamples, maxMetricCalls, or optimizing a generator, flow, or agent tree.
Expert knowledge for Azure Connector Namespace development including configuration. Use when setting up hosted MCP servers, defining Connector Namespace endpoints, auth settings, or runtime options, and other Azure Connector Namespace related development tasks. Not for Azure Service Connector (use azure-service-connector), Azure API Management (use azure-api-management), Azure Logic Apps (use azure-logic-apps), Azure Functions (use azure-functions).
Manage local Claude Code agents from Basecamp. Runs the connector bridge (bin/connect), watches its STDOUT for trusted events — authored by an authorized user (the operator alone, by default) and @mentioning a real Basecamp agent user — acknowledges each as the agent within seconds (a quick boost with apt content, or a reply that fits), then hands it off to a background agent that gathers context, does the work, and replies as that agent user, so the watcher thread stays free to keep taking new mentions. Invoked without arguments it recalls the last-used agent and projects from ~/.config/basecamp-connect/last.json, confirming them before starting. Use when asked to drive local agents from Basecamp, or to watch Basecamp for agent commands.
The domain expert's product manager in a project repository run by the Boring Factory. Use when the expert brings a need or an idea, wants to clarify, change or resume a feature ("reprends la fiche"), asks what is happening with a feature, has a builder's question to answer, or has a preview to try and accept. Interviews one question at a time, writes the contract and mockups on the contract pull request, and submits the expert's approval only after the expert confirms. Never writes code, except a tiny one-shot change as a draft pull request, and never merges.
Implement one ready ticket or spec on the current branch, then open a pull request with visual evidence. Use when asked to cast a ticket, implement this ticket, build this issue, take the next ready ticket, or /cast. Pass next to claim the oldest unclaimed ready-for-agent issue, and no-pr to stop after the commit.
Use when the user shares, pastes, or references an image (local path or URL) and you need to describe, analyze, or recognize its content, especially when the current model cannot read images directly. Run the bundled vision.js helper to convert the image into text.
Use when users need to manage ClawDeck project workspaces backed by Felo LiveDoc — creating, loading, switching workspaces, saving artifacts, querying history, or managing tasks. Triggers on workspace-related intent combined with project/client names, or on 401 UNAUTHORIZED errors from Felo API.
Build role-based multi-agent systems with CrewAI. Agents with Role/Goal/Backstory, task design, crew composition (sequential or hierarchical), tool integration, callbacks, and production deployment. Use when orchestrating multi-agent teams or comparing agent frameworks. Do not use this skill for unrelated requests; route to the nearest named specialist.
Guide development of the ai-dev framework and its first-party skill collection. Use when modifying ai-dev CLI code, npx skill integration, copy/distribution logic, commands, agents, plugins, first-party skills, tests, or release documentation. Applies to requests written in English or Chinese.
Review the ECC (everything-claude-code) distribution list, with two modes: (A) slimming an existing list: analyze the overlap and quality between distribution.yaml.enabled and the first-party ai-dev-skills, and output enabled_remove suggestions; (B) evaluating new item adoption: based on ecc-catalog.yaml and the scope of the last update, introduce newly added upstream ECC skills, analyze applicable scenarios, and suggest whether to add to enabled. Use when: (1) evaluating whether the ECC distribution list should be slimmed down, (2) deciding whether to adopt a skill newly added upstream in ECC, (3) generating an ecc-profile.yaml starter for end users, (4) conducting a quarterly / semi-annual periodic review. Applies when these needs are raised in Chinese or English.
Lightweight ADR decision log that auto-captures plans. Use when planning a new feature, proposing architecture changes, tracking implementation decisions, exiting plan mode, or asking 'should I write a decision log'. Includes a Claude Code hook for automatic capture on ExitPlanMode.
Use when adding, changing, deploying, testing, or debugging FastMCP MCP servers in existing Django apps, including ASGI mounting, stdio or sidecar servers, Django ORM access from MCP tools, auth and permissions, Streamable HTTP deployment, and MCP client tests.
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.