Skills
97,501 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Turn a complex, thought-through requirement into a detailed execution plan document plus a self-contained launcher prompt for a fresh agent session, and later review that session's report. Use when the user wants a plan before execution, wants execution handed to a separate session, or mentions 执行计划 / 计划文档 / 新会话执行 / 交接 prompt / 审查 report; also when reviewing a plan's execution report or dispatching a fix round. Grills what is still missing, applies ponytail taste, treats partial completion as legitimate, and never executes the plan itself. Prefer steady-do for implementation in this session and quick-do for clear, bounded changes.
Execute small, well-understood changes directly in the current session with minimal ceremony and ponytail taste. Use when explicitly requested, or automatically for localized fixes, mechanical edits, config tweaks, and short scripts with clear behavior and a known verification path. No plan documents or grilling; do not add tests by default. Prefer steady-do when design, compatibility, or acceptance decisions remain; prefer plan-brief when planning and execution need a separate-session handoff.
Guided setup assistant for self-hosting open-source Agenta and running an agent on it. Use when the user wants to self-host Agenta, set up an Agenta agent on their own machine or server, run an agent locally, or says "help me self-host Agenta", "set up an agent", "self-host and run an agent". Walks the user through the decisions, executes against the OSS self-host docs, and ends with a working, tested agent.
Conversation design skill for Salesforce Agentforce. Generates persona documents, topic architectures, instruction sets, utterance libraries, escalation matrices, and guardrail configurations. Validates existing agents against conversation design best practices with 120-point scoring.
Design new Claude skills from structured idea specifications. Use when the skill auto-generation pipeline needs to produce a Claude CLI prompt that creates a complete skill directory (SKILL.md, references, scripts, tests) following repository conventions.
Use when you need to audit, evaluate, govern, safely retire, or transactionally upgrade locally installed or distributed agent skills. Triggers include skill sprawl, stale, broken, or conflicting skills, local cleanup or quarantine, managed physical collections, pre-migration review, and periodic inventory checks.
Validate multi-skill workflows defined in CLAUDE.md by checking skill existence, inter-skill data contracts (JSON schema compatibility), file naming conventions, and handoff integrity. Use when adding new workflows, modifying skill outputs, or verifying pipeline health before release.
Universal skill diagnosis and optimization tool. Detect and fix skill execution issues including context explosion, long-tail forgetting, data flow disruption, and agent coordination failures. Supports Gemini CLI for deep analysis. Triggers on "skill tuning", "tune skill", "skill diagnosis", "optimize skill", "skill debug".
Explain and interpret a skill, a skills repository, or a skills system in a deep yet accessible teaching style. Use when the goal is to help readers truly understand how it works, why it works, what is worth learning, and how to design better skills.
Design and run a Spec-Driven Development (SDD) pipeline for AI software factories — where structured specifications are the input, AI agents generate the code, and quality gates enforce correctness at each phase: SPECIFY → DECOMPOSE → IMPLEMENT → VERIFY → DELIVER. Use when building or refining a spec-driven pipeline any AI coding tool (Claude Code, Cursor, Hermes Agent, Devin, OpenHands, droid) can follow, or when you need spec quality gates, phase-gate verdicts, NFR encoding, or format translation. Do not use for a single small change with a clear goal (classify first via bmad), for the control-plane protocol of intent contracts, autonomy gating, and failure routing around a pipeline (bmad), or for unvalidated problems (product-discovery).
Compose GitHub spec-kit, Ouroboros (ooo), and HKUDS CLI-Anything into one spec-driven delivery stack — Write → Freeze → Run, verified. Author the spec with `/speckit.*`, freeze it as an immutable ooo seed with machine-checkable acceptance criteria, arm the loop with agent-native CLI harnesses from CLI-Hub (`--json` output as evaluate evidence), and loop until verification passes. Routes three patterns — full-stack (spec-kit → ooo → cli-anything), loop-only (ooo), docs-only (spec-kit) — with explicit anti-patterns (two spec SSOTs, harness-first builds, seedless ralph loops). Use when the user wants spec-to-verified-artifact delivery, wants spec-kit and ooo to work together without fighting, or needs real software driven inside a verified loop. Triggers on: spec-stack, spec stack, write freeze run, spec to verified, speckit + ooo, ooo + cli-anything, spec-driven loop, verified delivery stack, seed with tools.
Multi-lane AI verification system. When the user says "run this through Urithiru", "Urithiru check", "full tower", or similar, fan the query out to 5-7 AI models simultaneously via OpenRouter, collect independent responses, and synthesize the best answer as the orchestrator. Antigravity is ALWAYS the orchestrator — the models are lanes (specialists), not the decision-maker.
Explicitly enable AGI mode; use when unsure how to move a problem forward or when self-evolution is needed. Consultations and queries are read-only. A general-purpose problem-solving assistant that resolves issues through evidence loops and continuously improves its methods and capabilities through retrospectives, evaluations and iterations
Create and manage isolated container sandboxes for AI agent code execution. Use when you need ephemeral environments to run untrusted code, execute agent workflows, or test in isolation. NOT for debugging existing apps (use troubleshooting skill).
Write a high-quality prompt for any LLM or AI assistant — Claude, Claude Code, ChatGPT, Gemini, Cursor, Windsurf, Copilot, or any coding / chat agent. Use this skill whenever the user asks to write, improve, refine, shorten, or rewrite a prompt; asks "how should I phrase this for [model]" or "what's a good prompt for [task]"; describes a task they want an AI to do but hasn't yet formulated it as a prompt; or pastes an existing prompt and asks for revision. Based on Boris's (Anthropic, Claude Code creator) prompt methodology — short and accurate prompts, plan-before-code, feedback loops, persistent context in files. The universal principles (short, plan-first, feedback-loop, no-padding) apply to any LLM; the Claude-Code-specific anchors (CLAUDE.md, @file, slash commands) only apply when the target is Claude Code. If the user's intent is unclear (target model, deliverable, scope, or whether the AI has a way to self-verify is missing), ask 1–3 targeted clarifying questions via AskUserQuestion before writing the prompt.
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design through one question at a time, then routes the outcome to write-idea, write-prd, write-techspec (refactors/bug fixes), or a direct task.
Use when running autonomous loops, repeated operations, or when detecting stagnation patterns - enforces rate limits, protects configuration files, manages recovery with cooldown periods, and prevents infinite loops during autonomous development
Optimize CLAUDE.md files using progressive disclosure: reduce bloat, move details to references, extract reusable patterns into skills, improve context efficiency. Use when CLAUDE.md is too long or needs restructuring.
Build applications that programmatically interact with GitHub Copilot. The SDK wraps the Copilot CLI via JSON-RPC, providing session management, custom tools, hooks, MCP server integration, and streaming across Node.js, Python, Go, and .NET.
Generate a project-local verification skill that drives your app the way a user does — any language, framework, or platform. Use for /create-verification-skill, "make a control skill for this repo", or when a project has no scripted way to prove UI/CLI/service behavior.
Use a headed browser to automatically complete DingTalk channel integration for QwenPaw. Applicable when the user mentions DingTalk, developer console, Client ID, Client Secret, bot, Stream mode, binding or configuring a channel. Supports pausing when a login page is detected and resuming after the user logs in.
Answers AI agent evaluation methodology questions with practical, opinionated guidance grounded primarily in Microsoft's agent evaluation ecosystem (MS Learn, Eval Scenario Library, Triage & Improvement Playbook, Eval Guidance Kit) supplemented by select industry sources.
Git Context Controller (GCC) v2 — Lean agent memory backed by real git. Stores hash + intent + optional decision notes instead of verbose markdown. Auto-bridges to aiyoucli vector memory when available. Dual mode: git-backed (lean index.yaml) or standalone (markdown fallback). Triggers on /gcc commands or natural language like 'commit this progress', 'branch to try an alternative', 'merge results', 'recover context'.
Google Agent Development Kit (ADK) for Python. Capabilities: AI agent building, multi-agent systems, workflow agents (sequential/parallel/loop), tool integration (Google Search, Code Execution), Vertex AI deployment, agent evaluation, human-in-the-loop flows. Actions: build, create, deploy, evaluate, orchestrate AI agents. Keywords: Google ADK, Agent Development Kit, AI agent, multi-agent system, LlmAgent, SequentialAgent, ParallelAgent, LoopAgent, tool integration, Google Search, Code Execution, Vertex AI, Cloud Run, agent evaluation, human-in-the-loop, agent orchestration, workflow agent, hierarchical coordination. Use when: building AI agents, creating multi-agent systems, implementing workflow pipelines, integrating LLM agents with tools, deploying to Vertex AI, evaluating agent performance, implementing approval flows.
Write less code and say less about it. Applies YAGNI and stdlib/native-first so the agent writes the minimum code that needs to exist, and responds tersely — stripping filler, hedging, and pleasantries while keeping code, identifiers, and technical terms exact. Use whenever writing, modifying, refactoring, reviewing, or explaining code, or any response where output volume drives token cost — even if the user never says "minimal" or "concise". Especially in agentic coding, where the volume of generated code and prose runs up the bill.
Generate and deploy a pro-code AI agent using LangGraph and SAP GenAI Hub, deploy it to SAP BTP Cloud Foundry, create a Joule capability with A2A (Agent-to-Agent) action, and deploy it to Joule. Supports TypeScript (Express or CAP) and Python. Use this skill whenever the user mentions: building a custom agent for Joule, deploying an AI agent to BTP Cloud Foundry, A2A protocol integration with Joule, pro-code agent extensibility, LangGraph agent on SAP BTP, connecting an external agent to Joule, "bring your own agent" for Joule, creating a Joule capability with A2A action, CAP agent, MTA deployment, or any combination of LangGraph/A2A/Joule/BTP/Cloud Foundry/CAP in an agent development context. Also trigger when the user wants to scaffold, modify, or redeploy an existing A2A agent project.
Self-evolving behavioral geometry and metacognitive lens for AI agents. Tracks perceptions, overrides, protections, self-observations, decisions, and curiosities that evolve from every experience. Use when the agent needs to learn from mistakes, develop self-awareness, track confidence in decisions, maintain behavioral guardrails from failures, preserve emergent behaviors, or cultivate active curiosities. Triggers on all conversations, corrections, errors, and reflective moments. Also use when setting up a new agent's self-learning system, configuring feedback loops, or packaging metacognitive capabilities.
Engineer effective LLM prompts using zero-shot, few-shot, chain-of-thought, and structured output techniques. Use when building LLM applications requiring reliable outputs, implementing RAG systems, creating AI agents, or optimizing prompt quality and cost. Covers OpenAI, Anthropic, and open-source models with multi-language examples (Python/TypeScript).
AI-assisted pair programming with multiple modes (driver/navigator/switch), real-time verification, quality monitoring, and comprehensive testing. Supports TDD, debugging, refactoring, and learning sessions. Features automatic role switching, continuous code review, security scanning, and performance optimization with truth-score verification.
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.