Skills
96,981 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Route goalflow (wanmol/goal-flow) work — a LangGraph framework that combines workflow graphs with agent loops — into exactly one mode: fit check, transpiling a Dify DSL export into runnable LangGraph Python, authoring workflow nodes and edges, building an `agent_kit` loop with middleware and a harness, wiring the serving layer (data adapters, SSE streaming, HITL, Redis/MySQL, API-key registration), or running the pre-publish security gate. Use when the user wants Dify's visual design without Dify's runtime, a graph node that hosts an agent loop, an OpenAI-compatible wire protocol over their own workflows, or prompt-injected `SKILL.md` capabilities. Triggers on: goalflow, goal-flow, dify to langgraph, dify transpiler, dify DSL export, BaseWorkflow, agent_kit, AgentBaseNode, DataAdapter, chunk processor, HITL interrupt, dify2langgraph. Route plain graph-API questions to `langgraph-fundamentals`.
AI Self-Improving Agent v2 - Learn from mistakes, corrections, and successes. Three-layer system: passive capture + proactive check + proactive skill generation (inspired by Hermes Agent). Avoid repeating the same errors, remember user preferences, and auto-generate reusable Skills from successful task completions.
Rewrite a memory or context file (CLAUDE.md, AGENTS.md, a todo or notes file) into Honey-terse form so it costs fewer input tokens every session, without losing meaning. Backs up the original first. Use when asked to shrink or compress context files, trim CLAUDE.md, or cut per-session input cost. Prose only — never code, config, or data.
Use when the user asks to “create a hook,” “add a PreToolUse/PostToolUse/Stop hook,” “validate tool use,” “implement prompt-based hooks,” “use ${CLAUDE_PLUGIN_ROOT},” “set up event-driven automation,” “block dangerous commands,” or mentions hook events (PreToolUse, PostToolUse, Stop, SubagentStop, SessionStart, SessionEnd, UserPromptSubmit, PreCompact, Notification). Provides comprehensive guidance for creating and implementing Claude Code plugin hooks, focusing on the advanced prompt-based hooks API
Change-size check run inside the tempa-spec schema's `proposal` phase, before the \"What Changes\" list is finalized. Detects when a draft bundles several independent changes that each deserve their own change, stops and asks the user whether to split (default: one change; the agent never splits on its own), and when the user splits, records the deferred slices in proposal.md's Out of Scope section, discovery.md's Key Decisions, and openspec/BACKLOG.md. Rationale is in docs/research/2026-08-27-mattpocock-skills-gate-and-seams.md.
Automate JIRA ticket processing through domain analysis, initialization, prompt evaluation-optimization and final task execution with comprehensive error handling in thinking model. Use when given a JIRA ticket URL (format: https://[domain].atlassian.net/browse/[TICKET-ID]) to process end-to-end.
Implement exactly one ready Phat task within its documented scope, update its Goal/Blocked by/Todo checklist with verified progress, preserve unrelated changes, apply the project's verification and commit policy, and prepare the task for review. Use when a task is todo or an existing in-progress continuation with all dependencies complete.
Iteratively improve any output until measurable criteria are met. Use when the user wants to refine existing work against specific standards — whether it's code, prose, data, config, or any other artifact. Triggers on phrases like "improve this", "make it better", "iterate", "refine", "keep improving", "not good enough yet", "optimize this", "polish this", "tighten this up", "ปรับปรุง", "ทำให้ดีขึ้น", "ยังไม่ดี", "แก้ให้ดีกว่านี้", "iterate ต่อ", or when the user provides criteria and wants repeated improvement until they're satisfied. Also use when the user gives feedback on output and expects you to keep refining, even if they don't say "improve" explicitly.
Use when the user mentions init-simple-memorix, too many or noisy Memorix hooks, project-level or global hooks configuration, missing Memorix MCP tools, full mode, WorkBuddy MCP startup failure, trust approval, Node argument compatibility, or re-trimming after setup or upgrade.
Meta-orchestration guide for choosing optimal reasoning patterns. Analyzes problem characteristics and recommends which cognitive methodology to use - tree-of-thoughts (find best), breadth-of-thought (explore all), self-reflecting-chain (sequential logic), or direct analysis. Use when facing complex problems and unsure which reasoning approach fits best.
AI-Native Issue-Driven development workflow. From GitHub Issue to merged PR: parse issue, explore codebase, design technical plan, execute with agent team, create PR, and cleanup. Use when a user wants to implement a GitHub Issue end-to-end: `/issue-flow #123` or `/issue-flow` to pick from open issues.
JavaScript/TypeScript SDK for inference.sh - run AI apps, build agents, integrate with all models. Package: @inferencesh/sdk (npm install). Full TypeScript support, streaming, file uploads. Build agents with template or ad-hoc patterns, tool builder API, skills, human approval. Use for: JavaScript integration, TypeScript, Node.js, React, Next.js, frontend apps. Triggers: javascript sdk, typescript sdk, npm install, node.js api, js client, react ai, next.js ai, frontend sdk, @inferencesh/sdk, typescript agent, browser sdk, js integration
Ingest a source into the knowledge bundle. Use when the user wants to capture, ingest, file, process, or "add this" — a note, transcript, email, PDF, image, web page, or any raw source — into a knowledge/ bundle, or drops content for processing. Reads the source once, extracts its signal, and integrates it across the bundle under the trust model so knowledge compounds instead of being re-derived per query.
Portable OKF knowledge bundle for coding-agent memory and LLM wikis. Use when the user wants to start a wiki, ingest a source, document a repository, query project knowledge with citations, lint for drift, or visualize connections; whenever a repo has a knowledge/ bundle that should inform the task; and when another kb-* skill needs the shared spec, glossary, templates, or trust model.
Use this skill when creating, naming, renaming, or organizing skills in this repository. Always use it when you need to decide rules for skill directory names, propose names that follow the existing convention, organize skills so related entries stay close together, or distinguish local skills from external ones.
An interactive course tutor for the AI Engineering from Scratch course. Reads LEARNING.md, finds the next lesson, teaches it section by section in the terminal, then quizzes you and records progress. Works in a cloned repository or entirely through raw.githubusercontent.com — no setup required. Trigger phrases: “下一课”, “教我”, “继续课程”, “我们来学习”, “继续学习”, or “next lesson”, “teach me”, “continue the course”, “let's learn”, “resume learning”.
Build voice AI agents with LiveKit Agents SDK. Use when the user asks to "build a voice agent", "create a LiveKit agent", "add voice AI", "implement handoffs", "structure agent workflows", or is working with LiveKit Agents SDK. Covers both LiveKit Cloud and self-hosted deployments using lk CLI.
Review a solution before approving it, like a GitHub changes tab: show a file tree of red and green pseudo-code for inline comments, then respond with Approve, Commented, or Rejected. Use when the user wants to see what would change, a git-like diff, or the PR as if it were ready.
Appends structured decision entries to the yahoo-mlb decision log (tracker/decisions-log.md) on behalf of any agent in the MLB team. Validates entries against the authoritative schema, serializes concurrent writes from parallel agents, and runs the Monday calibration pass to fill in outcomes and update the variant scoreboard. Use when any MLB agent needs to record a decision, when the coach requests "log decision", "append to decision log", "record variant outcome", or runs the "calibration pass".
Control, automate, and interact with Android and iOS phones via the Mobilerun API. Mobilerun automates mobile devices it is NOT a desktop browser automation tool. Use whenever the user wants to automate a task on a mobile device or run a cloud task on a phone. Use when: (1) tapping, swiping, typing, or navigating phone UI directly, (2) running autonomous AI agent tasks on a phone, (3) taking screenshots or reading screen state, (4) managing devices — provisioning, renaming, rebooting, terminating, (5) managing apps — install, uninstall, list packages, (6) configuring proxies, eSIM, GPS location, or file transfers, (7) managing credentials for automated app logins, (8) subscribing to task webhooks, (9) multi-device operations — fleet management, parallel tasks across phones. Do NOT use for desktop browser automation, computer screenshots, or controlling a desktop OS. Also load when the user mentions Mobilerun, Droidrun, or phone control. Supports both personal devices (via Portal APK) and cloud-hosted devices. Requires a Mobilerun API key (prefixed dr_sk_) and a connected device.
Guide users through configuring key MongoDB MCP server options. Use this skill when a user has the MongoDB MCP server installed but hasn’t configured the required environment variables, or when they ask about connecting to MongoDB/Atlas and don’t have the credentials set up
OpenRouter unified-LLM-API expertise. Use when code POSTs to `openrouter.ai/api/v1`, imports `@openrouter/sdk`, `openrouter` (Python), or `@openrouter/agent`; when the OpenAI SDK is pointed at `https://openrouter.ai/api/v1`; when the user asks about routing across hundreds of LLMs (OpenAI, Anthropic, Google, DeepSeek, Meta, xAI, Mistral, etc.) through one endpoint; when working with provider routing, model fallbacks (`models` array), the Auto Router (`openrouter/auto`), `:nitro`/`:floor`/`:online` model suffixes, presets (`@preset/...`), prompt caching with `cache_control`, structured outputs, tool calling, the `openrouter:web_search` server tool, reasoning tokens, multimodal (image/PDF/audio/video) inputs, BYOK provider keys, OAuth PKCE, the `openrouter:` plugins (`web`, `file-parser`, `response-healing`, `context-compression`), or generation/usage accounting via `/api/v1/generation` and `/api/v1/key`.
Main entry point for working with Opper (https://opper.ai), the AI gateway for agents. Use when the user mentions Opper without choosing a surface: "set up Opper", "try Opper", "build with Opper", "migrate to Opper", or "help me with Opper". Explore context and route to opper-mcp for agent-assisted setup, platform operations, and model tests; opper-cli for explicit CLI and shell workflows or launching coding agents through Opper; opper-api for application code (the OpenAI, Anthropic or Google SDKs and agent frameworks on the compat endpoints), raw HTTP, and platform concepts; or opper-multimodal for media, files, and realtime voice. Own setup, verification, and follow-up.
Run a fleet of Claude Code sessions on this machine and make them talk to each other with SendMessage — from Claude Code itself or through a real Claude relay when the caller is Codex. Launch them in cmux or herdr windows, panes and splits (auto-detected), address them correctly, hand out work, collect or recover replies, and tear them down cleanly when the user asks. Use whenever the user wants two or more Claude sessions working together, asks Codex to contact an existing Claude session, mentions SendMessage, ListAgents, peer sessions, cross-session messaging, or agents messaging each other; whenever they want sessions in new windows, workspaces, panes or splits; and especially when a send fails with "not an agent in this conversation", "re-send with the ref", or the caller lacks Claude's native messaging socket.
Manage Claude Code's global tool permission configuration, automatically adding MCP commands or other tools to allowedTools so manual approval is not needed on every use. Workflow: confirm the command the user wants to add -> confirm the scope (global ~/.claude.json by default) -> perform the addition -> verify and remind to restart.
This skill should be used when creating, validating, or converting Product Requirements Documents (PRDs) to LLM-native format. Use this skill when the user asks to write a PRD, review a PRD for compliance, convert an existing PRD to machine-readable format, or ensure PRD quality for AI agent consumption. This skill is optimized for users working with agentic coding frameworks who need strict requirement specifications to prevent context poisoning, hallucination, and mesa-optimization.
Prepare Sprint 0 (walking skeleton) for a project based on its Tech Design and ADRs: classify ADRs as cross-cutting or story-specific, define the route ownership map by role, write the skeleton's Definition of Done, propose a plan, and stop for human approval. Use when the LEAD (or user) needs to prepare, plan, or initialize a project's Sprint 0 / walking skeleton before any user story is built.
Runs autonomous loop fetching stories from GitHub Issues. Implements and closes issues as done. Triggers on "loop through my PRDs", "work on my issues", "start the autonomous loop", "implement my PRDs", or requests to work through GitHub issues autonomously.
Route requests to the focused Randroid development workflows and explain which capability to use. Use when the user asks what Randroid skills are available, invokes the Randroid namespace without naming a workflow, or needs help choosing among loop execution, PR feedback handling, VibeReview playtesting, visual de-slopping, and gameplay interaction or feedback work.
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.