Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Implement OpenAI Harness Engineering practices in any repository—AGENTS.md, PLANS.md, deterministic smoke/test/lint harness commands, strict architecture boundaries, observability from day one, and entropy-control audits for reliable autonomous agent runs.
Delegate complex, long-running tasks to Manus AI agent for autonomous execution. Use when user says 'use manus', 'delegate to manus', 'send to manus', 'have manus do', 'ask manus', 'check manus sessions', or when tasks require deep web research, market analysis, product comparisons, stock analysis, competitive research, document generation, data analysis, or multi-step workflows that benefit from autonomous agent execution with parallel processing.
Explore requirements and approaches through collaborative dialogue before writing a right-sized requirements document and planning implementation. Use for feature ideas, problem framing, when the user says “let's brainstorm”, or when they want to think through options before deciding what to build. Also use when a user describes a vague or ambitious feature request, asks “what should we build”, “help me think through X”, presents a problem with multiple valid solutions, or seems unsure about scope or direction — even if they don't explicitly ask to brainstorm.
Automatically discover life science APIs online, create ToolUniverse tools, validate them, and prepare integration PRs. Performs gap analysis to identify missing tool categories, web searches for APIs, automated tool creation using devtu-create-tool patterns, validation with devtu-fix-tool, and git workflow management. Use when expanding ToolUniverse coverage, adding new API integrations, or systematically discovering scientific resources.
Use KERNEL managed auth or vault credentials after choosing a browser path that genuinely requires an authenticated website ui or authenticated browser behavior. prefer a dedicated api, cli, mcp server, or authoritative local records for operations they support; a url alone does not require browser auth. reuse or create auth connections, complete hosted login without abandoning the task, launch profile-backed browsers with telemetry, and handle reauthentication safely. for per-user vault credentials, let the user choose KERNEL-hosted collection or 1Password brokered approval, and fill only with explicit approval.
Use when the agent wants to define, list, inspect, or execute GUI macros via the MacroCLI CLI. Macros are parameterized, CLI-callable workflows — the agent invokes `macro run <name>` and the system handles backend routing (plugin, file transform, accessibility, compiled GUI replay).
Build, test, publish, invoke, schedule, and update Railcode managed agents with the Railcode CLI. Use when creating an organization or personal managed agent, editing an agent manifest (JSON or YAML), running a draft or saved agent, investigating a run, managing its cron schedule, running it from Slack (@Railcode $agent), pairing it with a companion app, processing files in its sandbox, or acting through a connector one person owns (their Gmail, Slack, Granola, ...). Do not use for static Railcode apps, in-app LLM tool loops (llm.generate({ tools }) — see create-railcode-app), or general organization administration.
Delegate a coding task to Muse Code (`muse`) as a background implementer, then review its diff and land it yourself. Use this whenever the user wants to hand implementation work to Muse — phrasings like "have Muse do X", "delegate this to Muse", "run it through Muse", or "use Muse to implement/fix/refactor" — or wants to run a queue of coding tasks through Muse while staying the reviewer. Prefer it when the user will review the diff and commit it themselves. It applies only when the user wants that separate `muse exec` implementer run: if you are Muse yourself and the user simply asks you to write the code, do it directly. DO NOT USE for tasks small enough to do inline.
Ultra-compressed Thai+English communication mode. Cuts ~60-75% of output tokens by speaking simple Thai while preserving technical accuracy. Triggers: "/pordee" / "/pordee full" / "พอดี" / "พอดีโหมด" / "พูดสั้นๆ" → enable full - "/pordee lite" → enable lite - "/pordee stop" / "หยุดพอดี" / "พูดปกติ" → disable - "/pordee-stats" / "/pordee-stats --share" → show token usage stats
Execute implementation tasks from design documents using markdown checkboxes. Use when (1) implementing features from feature-design-assistant output, (2) resuming interrupted work, (3) batch executing tasks. Triggers on 'start implementation', 'run tasks', 'resume'.
Create, generate, update, repair, trim, or split minimal AGENTS.md files for repository roots and confirmed nested modules using progressive disclosure. Use for missing, bloated, contradictory, or outdated agent instructions, monorepos, services, packages, module context, and portable agent guidance without assuming any fixed agent folder structure.
Make 100 versions of Claude fight to the death over one task. Spins up N sub-agents (default 100, --quick for 16), gives every one the exact same task plus a different strategy card (reasoning mode, workflow, strategy), then runs a single-elimination bracket: they attack each other's solutions, defend and revise, and a judge scores every match on a written rubric until one solution survives. Use when the user is not satisfied with an answer, calls it a bad answer, says try again or do better, says "arena", or asks to make them compete
Create structured plans for any multi-step task -- software features, research workflows, events, study plans, or any goal that benefits from structured breakdown. Also deepen existing plans with interactive review of sub-agent findings. Use for plan creation when the user says 'plan this', 'create a plan', 'write a tech plan', 'plan the implementation', 'how should we build', 'what's the approach for', 'break this down', 'plan a trip', 'create a study plan', or when a brainstorm/requirements document is ready for planning. Use for plan deepening when the user says 'deepen the plan', 'deepen my plan', 'deepening pass', or uses 'deepen' in reference to a plan.
Automated issue loop with checkpoint/resume: fetch open GitHub issues → dependency-aware topological sort → implement each issue end-to-end → review with /review-it → ship with /ship-it → repeat. Persists state to .loop-state.json for crash recovery. Triggers on: loop-it, loop issues, auto implement, 批量实现, 循环实现, 实现所有issue, 恢复循环, resume loop
Synapse A2A agent communication -- sending messages, spawning agents, delegating tasks, sharing memory, managing the LLM wiki, and coordinating file edits. Use this skill when: running synapse send/reply/broadcast/interrupt, spawning agents with synapse spawn or synapse team start, sharing knowledge with synapse memory, managing wiki pages with synapse wiki, locking files with synapse file-safety, checking agent status with synapse list/status, or orchestrating any multi-agent workflow. For AI/programmatic use, prefer synapse list --json, synapse status <target> --json, or the MCP list_agents tool instead of interactive synapse list.
Browser automation for AI agents via inference.sh. Navigate web pages, interact with elements using @e refs, take screenshots, record video. Capabilities: web scraping, form filling, clicking, typing, drag-drop, file upload, JavaScript execution. Use for: web automation, data extraction, testing, agent browsing, research. Triggers: browser, web automation, scrape, navigate, click, fill form, screenshot, browse web, playwright, headless browser, web agent, surf internet, record video
Routing-first visual approval gate for AI agent plans, markdown specs, and diffs. Use when a human needs to review a concrete plan before execution, inspect a targeted diff in a browser, mark up a spec/PRD/architecture note, or set up the review loop on Claude Code, Gemini CLI, Codex CLI, or OpenCode. Route planning/spec creation to `task-planning` or `ralph`, broad PR-policy review to `code-review`, rendered-UI critique to `agentation`, and fresh-session browser verification to `browser-harness`.
Track progress across sessions using SESSION.md with git checkpoints and concrete next actions. Converts IMPLEMENTATION_PHASES.md into trackable session state. Use when: resuming work after context clears, managing multi-phase implementations, or troubleshooting lost context.
Luban — a Skill workshop. Refine a usable Skill into a public asset that can be understood, installed, shared, validated, and continuously improved. Its craft-based method has five steps: inspect the material (challenge the Skill’s premise first and say plainly when it is not worth refining), survey the field (search online for similar Skills and understand its place in the ecosystem), measure it (assess structure, real-world tests, and live runs together—the latter means checking actual outputs, since green CI can lie), plane slowly (freeze the original as a baseline; keep changes only if they pass validation, otherwise revert; wherever possible, preserve validation methods as repository tools and conventions), and rework (publication is not the finish line; keep a comparison watchlist and use real feedback in the next round). Use when users want to upgrade, optimize, refine, productize, or publish a Skill they developed. The final deliverables are a structured “Skill Refinement Report,” rewrite snippets ready to replace the originals, and a shareable “certificate of mastery” result card. Triggers include, but are not limited to: “let Luban review this skill,” “refine my skill,” “upgrade my skill,” “optimize this skill,” “skill checkup,” “skill audit,” “productize my skill,” “how do I publish this skill,” “compare it with similar skills,” “why is nobody installing my skill,” “help me publish my skill to GitHub/ClawHub,” and “improve SKILL.md.” Even if the user only shares a Skill directory, GitHub repository link, or SKILL.md and asks “how should I change this?”, trigger it whenever the context is to make the Skill more useful and shareable. Do not use it to create a new Skill from scratch (use skill-creator), for ordinary code review (use code-review), or to rewrite a general prompt unrelated to Skill assets.
Develop AI agents, tools, and workflows with Mastra v1 Beta and Hono servers. This skill should be used when creating Mastra agents, defining tools with Zod schemas, building workflows with step data flow, setting up Hono API servers with Mastra adapters, or implementing agent networks. Keywords: mastra, hono, agent, tool, workflow, AI, LLM, typescript, API, MCP.
Orchestrates BMAD workflows for structured AI-driven development. Use when initializing BMAD in projects, checking workflow status, or routing between 4 phases (Analysis, Planning, Solutioning, Implementation). Manages project configs, tracks progress through project levels 0-4, and coordinates with specialized workflows. Trigger on /workflow-init, /workflow-status, or when users need BMAD setup
Design, build, run, and test Restate durable services, virtual objects, workflows, and AI agents across TypeScript, Python, Java, and Go. This skill should be used when the user mentions "restate", "durable execution", "virtual object", "restate service", "restate workflow", or "durable agent" or wants to build resilient backend services, AI agents, or workflows with automatic failure recovery. Also use when converting existing applications or migrating from workflow orchestrators to Restate. Use proactively when a project contains restate dependencies in package.json, requirements.txt, pyproject.toml, pom.xml, build.gradle, or go.mod.
Fix or change an AI agent's behavior using MLflow traces and regression tests. Use for wrong or incomplete answers, hallucinations, missing tool calls, new business rules, trace-driven debugging, or context and token-cost optimization. Diagnose the actual failure, plan a test, implement the smallest fix, and verify the new behavior without breaking existing cases.
This skill should be used when the user asks about "Make MCP server", "Make MCP tools", "MCP token", "Make OAuth", "scenario as tool", "MCP scopes", "Make API access", "connect Make to Claude", "scenario not appearing", "MCP timeout", "MCP connection refused", or discusses configuring, troubleshooting, or understanding the Make.com MCP server integration. Provides technical reference for connection methods, scopes, access control, and troubleshooting.
Guide users through configuring ZenMux Base URL, API endpoint, API Key, and model settings for tools or SDKs. Use when the user wants to set up, configure, or connect ZenMux in Cursor, Claude Code, Cline, Cherry Studio, Open-WebUI, Dify, Obsidian, Sider, Copilot, Codex, Gemini CLI, opencode, or custom SDK code. Trigger on "configure", "setup", "set up", "base url", "endpoint", "api key", "how do I set up ZenMux", "help me fill in API settings", "接入", "配置", "设置", "base url 填什么", "怎么填", "怎么接入", "API 地址", "接口地址". Treat the user as a first-time user and guide step by step. Do not use for usage queries or docs lookups; use zenmux-usage or zenmux-context instead.
Entry point for AI-agent work on AWS: pick a runtime, plan a migration for existing workloads, and build an executable POC — one phased flow. Triggers on: which runtime for my agent, AgentCore vs ECS vs EKS vs Lambda, AgentCore vs Lambda MicroVMs, deploy an AI agent on AWS, agent architecture on AWS, I have an agent idea what do I build, move/migrate my agents to AWS, agent migration plan, add AgentCore services (memory, gateway, identity, policy, observability) to an agent already on AWS, Temporal on AWS (migrate/run Temporal workers on AWS, a service orchestrated by Temporal, Temporal Cloud vs self-hosted). Temporal Workflow code is never rewritten into Step Functions. Requires at least one agentic component — a purely non-agent system (plain services, batch jobs, HTTP endpoints, non-agent Temporal Activities) is out of scope, redirected to gcp-to-aws / heroku-to-aws / llm-to-bedrock. Not for: compute/data migration with no AI agent; pure LLM SDK rewrite (use llm-to-bedrock); per-model pricing.
/c-level-agents:cross-eval [memo] — Multi-model consensus on a board memo or strategy brief. Claude + Codex + Gemini cross-review with graceful degradation. Use when a high-stakes memo needs an independent sanity check before the boardroom — e.g. a bet-the-company pivot or fundraise terms.
Adversarial review (对抗 review) of a finished deliverable — a change ready for a PR, an investigation report, a recommendation, a batch of record edits — before it is reported to the user. Spawns fresh reviewer subagents briefed to disprove it from three directions (open, over-engineering, residue), verifies and fixes what holds, then reports. Run once per deliverable when it is considered done, and on request at level strong (强力) or max (极致). Not for answers to questions, lookups, or steps inside a task.
This skill should be used when executing Bankr requests, submitting prompts to Bankr API, polling for job status, checking job progress, using Bankr MCP tools, or understanding the submit-poll-complete workflow pattern. Provides the core asynchronous job pattern for all Bankr API operations.
Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to “open a website”, “fill out a form”, “click a button”, “take a screenshot”, “scrape data from a page”, “test this web app”, “login to a site”, “automate browser actions”, or any task requiring programmatic web interaction.
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.