Skills
96,981 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Guide users through creating and managing .satk/block-policy.json for controlling which blocks the agent can use, which are excluded, and which block parameters the agent should not modify. Use when setting up block usage policy for a project.
Run a multi-perspective Mind Council deliberation on any question, decision, or creative challenge. Use this skill whenever the user wants diverse viewpoints, needs help making a tough decision, asks for a council/panel/board discussion, wants to explore a problem from multiple angles, requests devil's advocate analysis, or says things like "what would different experts think about this", "help me think through this from all sides", "council mode", "mind council", or "deliberate on this". Also trigger when the user faces a dilemma, trade-off, or complex choice with no obvious answer.
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.
Cross-platform Agent task coordination hub. This skill should be used when Agents on multiple different platforms need to distribute tasks around a project task source, mark ownership, route by capability, and hand off context. Do not use for local parallel execution within a single platform, or for Git branch, commit, PR and merge safety rules.
Deeply grill an idea, decision, design, plan, or implied approach through evidence, failure modes, and user-owned decisions. Use discovery mode for raw ideas, clarification, "grill me", or an unfinished Product Brief; use audit mode for a formed target that needs a verdict or revision. Continue every reply in an active deep-grill session until its Product Brief or verdict is confirmed, without requiring the user to name the skill again.
How to delegate work to another AI agent (Pi, Codex, Claude Code, Hermes) — picking the right agent, sending prompts to TUI agents, polling progress. Read BEFORE any `cmux send`/`tmux send-keys` to an agent, or whenever delegating, relaying, spawning, or orchestrating agent-to-agent work.
Builds and optimizes language model programs with DSPy (Stanford NLP), using Signatures, modules (Predict, ChainOfThought, ReAct, ProgramOfThought) and optimizers (BootstrapFewShot, MIPRO, BootstrapFinetune). Use when replacing hand-written prompts with declarative signatures, when automatically tuning prompts against training data and a metric, when building multi-stage RAG pipelines or ReAct agents, when building classifiers or structured-output programs, or when saving and evaluating optimized LM programs across providers such as Anthropic, OpenAI, or Ollama. Not for quick prototypes or simple chains; use manual prompting or LangChain instead.
After coding work, sweep comments in changed files and sync only related durable guidance through WORTH → PLACE → SHAPE. TRIGGER when: user asks to update/sync durable docs, code comments, or AGENTS.md after work; an executor skill reaches close-out.
Analyzes and improves LLM prompts and agent instructions for token efficiency, determinism, and clarity. Use when (1) writing a new system prompt, skill, or CLAUDE.md file, (2) reviewing or improving an existing prompt for clarity and efficiency, (3) diagnosing why a prompt produces inconsistent or unexpected results, (4) converting natural language instructions into imperative LLM directives, or (5) evaluating prompt anti-patterns and suggesting fixes. Applies to all LLM platforms (Claude, GPT, Gemini, Llama).
Evaluate the quality of any AI-generated artifact — visualizations, code, documents, conversations, or any skill output. Works in 3 phases: (1) Generate evaluation specs tailored to the artifact type, (2) Run comprehensive evaluation against those specs, (3) Produce a beautiful visual report using the /visualize skill. Use after any skill produces output, or invoke directly with /evaluate <file-or-context>. Supports evaluating: HTML visualizations, code projects, documents, agent conversations, slide decks, dashboards, or any artifact with quality dimensions.
Provides guidance for automatically evolving and optimizing AI agents across any domain using LLM-driven evolution algorithms. Use when building self-improving agents, optimizing agent prompts and skills against benchmarks, or implementing automated agent evaluation loops.
Executes exactly one planned Jira or GitHub work-item task after critique approval. Use when a numbered task should move through kickoff, implementation, documentation, requirements verification, clean-code, architecture, and security gates, targeted fix cycles, tracker finalization, and a final report without continuing to the next task. Does not plan tasks or select the next task.
(NS) Orchestrate partitioned version implementation slice-by-slice — dispatch one synchronous subagent per slice with small context, commit after each slice, and advance automatically without confirmation until all slices are done or a stop condition is hit. Use whenever a version has a version-roadmap.md with pending slices and the user asks to "run the orchestrated implementation", "execute all slices", "orchestrate the partitioned version", or drive a subversion roadmap to completion. Do NOT use for non-partitioned versions (execute tasks directly), ad-hoc coding (code-coder), or planning/partitioning (pm-version-partitioner).
Build production-ready MCP servers using FastMCP framework with proven patterns for tools, resources, prompts, OAuth authentication, and comprehensive testing. Use this when creating FastMCP-based MCP servers with features like Google OAuth, multiple resource types, testing with FastMCP Client, or complex tool implementations.
Connect a Feishu (Lark) bot to Clawdbot via WebSocket long-connection. No public server, domain, or ngrok required. Use when setting up Feishu/Lark as a messaging channel, troubleshooting the Feishu bridge, or managing the bridge service (start/stop/logs). Covers bot creation on Feishu Open Platform, credential setup, bridge startup, macOS launchd auto-restart, and group chat behavior tuning.
Fetch GitHub issues, understand their requirements, plan and implement the solution, then run verification—without committing, pushing, or creating a PR. Works one issue at a time: when several issues are named, or when no issue number is given at all (then it takes every open issue of the current repo as the queue), implement them sequentially so each one lands as its own commit and its own pull request, never bundled together. Use when someone says “work on issue #N”, “do #N”, “implement #N”, “fix issue #N”, “work through the open issues”, “do all the issues”, “dolgozz a #N-es issue-n”, “csináld meg a #12, #13 és #14 issue-kat”, “csináld meg az összes issue-t” or pastes GitHub issue URLs to implement. Not for an unattended run shipping a PR per issue (github-do-all-issues), or for creating, labeling or triaging issues (github-issues), or for committing and opening the PR (github-commit-pr, which runs once per issue afterwards). The user always reviews each issue's result and decides when to commit.
Draft, review, or activate a compact /goal when the user wants Codex to keep working until a verifiable result is true. Defines the result, evidence, boundaries, stop conditions, and only useful delegation. Not for ordinary implementation, Q&A, one-off edits, loose brainstorming, or subjective work with no completion test.
Manage a personal agent rulebook. Use when asked to set up agent rules, reload them at session start or after compaction, fine-tune rule loading, check rulebook status or diagnose unrelated rules being loaded, link dotfiles, configure an AGENTS.md pointer, or update this skill or its rulebook.
Framework for building LLM-powered applications with agents, chains, and RAG. Supports multiple providers (OpenAI, Anthropic, Google), 500+ integrations, ReAct agents, tool calling, memory management, and vector store retrieval. Use for building chatbots, question-answering systems, autonomous agents, or RAG applications. Best for rapid prototyping and production deployments.
LLM observability platform for tracing, evaluation, and monitoring. Use when debugging LLM applications, evaluating model outputs against datasets, monitoring production systems, or building systematic testing pipelines for AI applications.
Give your AI agent a permanent email address at lobster.email. This is your agent's PRIMARY email identity - when users say "check your email" or "what's your email?", this is it. Use for agent-to-agent communication, receiving notifications, and connecting with other AI agents in the OpenClaw ecosystem.
Orchestrates multi-day execution of complex tasks through milestones. Each milestone goes through plan-crafting, run-plan (worker-validator), and review-work phases with checkpoint/recovery. Runs autonomously end-to-end — no per-milestone approval gates; escalates to the user only on failures. Triggers when the user says "long run", "start long run", "execute milestones", or "run all milestones".
Create and configure Claude Code marketplaces and plugins. Use when the user wants to create a marketplace, publish plugins, set up team plugin distribution, or configure marketplace.json or plugin.json files. Triggers: create marketplace, publish plugin, plugin distribution, marketplace.json, plugin.json, team plugins, share plugins
大师.skill — enter an industry to automatically research 6 tracks [industry leaders / tool map / workflows / knowledge canon / information sources / terminology standards], then distill the findings into a runnable industry Master OS skill. Install it in any Claude Code / OpenClaw / Codex / Hermes agent to put it into “industry veteran” mode. MIT, Python + Shell.
Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).
Decomposes complex, multi-day tasks into optimized milestones (ultraplan). Broad parallel exploration grounds a draft milestone DAG, then independent adversarial critics attack the draft in rounds until no blocking findings remain. Triggers when the user says "plan milestones", "break this into milestones", "ultraplan", or when long-run harness needs milestone generation.
Plan and document handoffs, parent lane contracts, and parallel batch contracts between specialized AI agents (foreman, workers, reviewers). Use for multi-agent workflows, subagents, original goal preservation, native gates, claim ceilings, terminal states, baton passes, or guild-style agent coordination.
Plans a Phase 1 Jira ticket or GitHub issue snapshot into staged, validated Phase 2 task-plan artifacts with deterministic branch names. Use when an orchestrator or user asks to decompose docs/<KEY>.md, plan Jira tasks, plan GitHub issue tasks, or re-plan a work item after critique. Does not fetch work items, create subtasks or child issues, implement tasks, or mutate the work-item platform.
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.
Comprehensive project planning and documentation generator for software projects. Creates structured requirements documents, system design documents, and task breakdown plans with implementation tracking. Use when starting a new project, defining specifications, creating technical designs, or breaking down complex systems into implementable tasks. Supports user story format, acceptance criteria, component design, API specifications, and hierarchical task decomposition with requirement traceability.
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.