Skills
96,981 packaged capabilities and playbooks for agents. Install any of them with one command.
Development
Use when the user wants the LLM to do its own ML research: a fully-autonomous loop that hacks the training code, runs it, and keeps changes that lower a single scalar metric (e.g. val_bpb). One agent proposes one change at a time, runs training in the user's env, keeps it only if the metric improves (advancing a git branch) else reverts, and loops forever until the human interrupts. A faithful adaptation of Karpathy's autoresearch. Not for the analysis-first variant that profiles before editing (that is ml-autoresearch), and not for a budgeted, plateau-stopping refactor.
Drive engineering work autonomously from intent to a verified terminal state. Picks a route (how much process) and a playbook (what kind of work), applies engineering principles, delegates across models, reviews independently, opens and watches PRs, lands them when granted, and closes out. Use for /kickoff or $kickoff, rigorous engineering work, or any change requested as shipped end to end.
Supports standalone and Seneschal-nested delivery with per-run state, inherited initiative contracts, and brokered decisions. Discovery-gated artifact-first orchestrator for compound-engineering product delivery. Resolves roadmap/readiness generation, runs brainstorm/plan/document-review loops, derives mergeable work packages with focused review units, executes each review unit through resolved work/code-review roles, and hands shipping to krt-release-marshal with CI break-prevention evidence. Use when turning an existing documented software project into a sequenced roadmap and PR/Jira delivery program. Runtime aliases may expose this as krt:compound-master.
Orchestrate the full delivery flow for the current project repository: direct krt-gitflow-knight for clean commits, krt-rebase-smith for clean branch history, resolve Jira Cloud or Server/Data Center explicitly, then open a GitHub pull request with bidirectional Jira/PR links. Use when the user asks to create/open a PR, prepare a pull request, ship current work, publish branch changes for review, or run the full gitflow + rebase + Jira + PR workflow. Runtime aliases may expose this as krt:release-marshal.
Ethereum Layer 2 landscape — Arbitrum, Optimism, Base, zkSync, Scroll, Unichain, Celo, Robinhood Chain, and more. How they work, how to deploy on them, how to bridge, when to use which. Includes per-chain DeFi ecosystems and critical corrections. Use when choosing an L2, deploying cross-chain, or when a user asks about Ethereum scaling.
Use this skill for ANY LangSmith/LangChain observability question. Covers two topics: (1) Adding tracing to your application (LangChain/LangGraph or vanilla Python/TS with @traceable), and (2) Querying traces for debugging, analyzing execution flow, and exporting trace data.
Real-time source of truth for software and model versions. Activate when adding dependencies, installing packages, or identifying Gemini model names. Bypasses internal knowledge cutoffs by querying live registries (NPM, PyPI, Go Proxy) and official documentation.
Configure LaunchDarkly Factory settings (GitHub App auto-flagging and auto-releasing) via the hosted MCP, or diagnose why a pull request was not classified / auto-flagged. Use when the user wants to turn on auto-flagging, map a GitHub repo to a LaunchDarkly project, change Factory account defaults, unmap a repo, or asks why Factory did not classify their PR
Dynamic lighting and environment in Decentraland scenes. LightSource (point and spot lights), shadows, SkyboxTime (day/night cycle), realm detection, and emissive materials for glow effects. Use when the user wants lights, shadows, skybox control, day-night cycle, or glowing materials. Do NOT use for PBR material properties like metallic/roughness (see advanced-rendering).
Build confidential smart contracts and dApps on EVM with Inco's TEE-based confidential computing: encrypted types (euint256/ebool/eaddress), encrypted ops (add/sub/mul/select/eq/rand), access control (e.allow), and attestation verification. Use for @inco/lightning Solidity + @inco/lightning-js SDK encrypt/decrypt, Foundry/Hardhat with a local covalidator or Base Sepolia, create-inco-app scaffolding, and confidential tokens, auctions, voting, or lottery dApps. Also covers confidential / hidden-information GAMES — deciding what stays private and which Inco feature goes where, then building fast: casino, cards, board, sealed auction, social deduction, fog-of-war, word/code-guessing. TRIGGER: imports "@inco/lightning" or "@inco/lightning-js" (or legacy "@inco/js"), mentions Inco, confidential EVM contracts, encrypted types, "what should be private in my game", on-chain poker/mafia/minesweeper/word-guessing, fog of war, provably fair. NOT for: ZK proofs, FHE/fhevm circuits.
Read and manage Linear issues, projects, cycles, teams, comments, and labels. Use this whenever the user wants to list their issues, search issues, create or update an issue, move an issue between states, add a comment, check a project or cycle, look up a team, or ask "what's on my plate in Linear" — even if they don't say "API" or "GraphQL". Also use it for any linear.app URL or an issue identifier like "ENG-123". Always start from this skill when interacting with this service — its bundled scripts and recipes are the fastest path.
Sync, audit, or rescue an existing Linear project against repo reality. Use when the user asks to sync Linear, reconcile tracking, audit this project, fix a messy board, revive a parked project, or score readiness with the rubric. Runs discovery + confirmation before writes; never silent reorganize. NOT for first-time bootstrap (use linear-setup) or one-off CRUD.
Implement Litestar authentication with custom authentication middleware, built-in security backends, JWT and session flows, route inclusion and exclusion rules, and typed auth context on `Request` / `ASGIConnection`. Use when establishing identity, issuing or validating credentials, or attaching authenticated user context in Litestar. Do not use for generic request parsing, broad security audits, or unrelated transport concerns.
Design and apply Litestar middleware for cross-cutting concerns such as CORS, CSRF, allowed-host checks, compression, rate limiting, logging, sessions, request enrichment, policy enforcement, and custom ASGI pipeline control. Use when behavior must wrap broad route sets consistently across the ASGI stack. Do not use for route-specific business rules, simple response mutation better handled by lifecycle hooks, or auth/guard policy work that belongs in security-focused skills.
Lumen — project knowledge keeper. Trigger on any /lumen command, on mentions of "lumen", "knowledge base", "wiki", "ADR", "document the project", on broad architectural questions about a codebase ("how does X work", "why did we choose Y"), or when the user pastes a transcript/spec to capture for future sessions.
Use on a brownfield codebase that needs a CONTEXT.md hierarchy. After intake, reverse-engineers the system and component architecture from existing code to produce root and component CONTEXT.md files. Use when that context would materially help; architect and triage do not require it. Also use when CONTEXT.md files are severely out of date.
Add concise, useful JSDoc where IDE hover help or a generated API reference needs a non-obvious contract. Use for exported JavaScript or TypeScript APIs, components, hooks, classes, complex types, or package publishing. Not for internal code, narrating comments, prose docs or behaviour changes
Build stateful AI agents and agentic workflows with LangGraph in Python. Covers tool-using agents with LLM-tool loops, branching workflows, conversation memory, human-in-the-loop oversight, and production monitoring. Use when - (1) building agents that use tools and loop until task complete, (2) creating multi-step workflows with conditional branches, (3) adding persistence/memory across turns with checkpointers, (4) implementing human approval with interrupt(), (5) debugging via time-travel or LangSmith. Covers StateGraph, nodes, edges, add_conditional_edges, MessagesState, thread_id, Command objects, and ToolMessage handling. Examples include chatbots, calculator agents, and structured workflows.
Discover OPC UA servers and create client connections in MATLAB using opcuaserverinfo, opcua, connect, setSecurityModel, and certificate trust functions. Use when discovering OPC UA servers on the network, connecting to OPC UA servers, authenticating with username/password or certificates, configuring security modes, handling certificate trust errors, fixing hostname mismatch warnings, troubleshooting connection failures or empty discovery results, or inspecting an OPC UA certificate (.der or .pem) for compliance issues. Trigger on: opcuaserverinfo, OPC UA discovery, find OPC UA servers, LDS setup, opcua, opc.ua.Client, connect OPC UA, OPC UA client, OPC UA security, OPC UA certificate trust, OPC UA certificate inspection, setSecurityModel, opc.ua.trustServerCertificate, opc.ua.exportClientCertificate, Industrial Communication Toolbox, OPC UA server connection, OPC UA server discovery.
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).
Comprehensive guide for creating software diagrams using Mermaid syntax. Use when users need to create, visualize, or document software through diagrams including class diagrams (domain modeling, object-oriented design), sequence diagrams (application flows, API interactions, code execution), flowcharts (processes, algorithms, user journeys), entity relationship diagrams (database schemas), C4 architecture diagrams (system context, containers, components), state diagrams, git graphs, pie charts, gantt charts, or any other diagram type. Triggers include requests to "diagram", "visualize", "model", "map out", "show the flow", or when explaining system architecture, database design, code structure, or user/application flows.
Safe migration workflows for EF Core database migrations, .NET version upgrades, and NuGet dependency updates. Includes rollback strategies and verification steps. Load when: "migration", "add migration", "ef migration", "update database", "upgrade nuget", "update packages", "dependency update", "version upgrade".
Refactor a growing WeChat Mini Program whose top-level navigation has become scattered across tabs such as home, profile, inbox, integrations, or settings into a clearer hub or center structure. Use when Codex needs to regroup top-level tabs, separate high-frequency action flows from low-frequency settings, define internal hub sections, or migrate detailed pages without collapsing everything into one long page.
Use host-side WeChat DevTools automation to inspect GUI-only runtime and interaction failures that do not show up in CLI `preview`. Trigger when page entry, taps, websocket-based DevTools automation, or run-directory evidence such as `report.json` and `trace.log` are needed to separate repo bugs from IDE session problems, local service blockers, or remaining visual-only issues.
Next.js 14+ App Router. Covers Server Components, routing, data fetching, caching, and deployment. Use when building Next.js applications with the App Router. USE WHEN: user mentions "Next.js", "App Router", "Server Components", "RSC", asks about "server actions", "Next.js routing", "Next.js caching", "Next.js deployment", "streaming", "Suspense in Next.js" DO NOT USE FOR: React (non-Next.js) - use `frontend-react` instead; Nuxt - use `nuxt3` instead; Remix - use `remix` instead; Pages Router (old Next.js) - consult KB for migration
Monitor and debug web and Node.js applications using OpenTelemetry, console logs, network requests, and distributed tracing. Use when the user asks about distributed tracing, correlating frontend/backend requests, OpenTelemetry, Jaeger, or monitoring application behavior.
Use when an OpenHarmony C++ change must be checked for call-chain completeness, especially for data propagation, IPC/proxy/stub paths, virtual overrides, callbacks, or dlopen/dlsym boundaries. Produces evidence tables and modification coverage matrices; the helper script only discovers candidate edges.
Use when measuring RAM/ROM impact of OpenHarmony multimodal input (MMI) subsystem changes on DAYU200 (RK3568). Covers ROM measurement via readelf section analysis and file size comparison, RAM measurement via /proc/PID/smaps, smaps_rollup, hidumper --mem, and baseline/current .so swap methodology. Includes code-level theoretical analysis for heap-allocated data structures.
Guidance for contributing code to open source projects the contributor does not maintain. Covers the fork and feature-branch workflow, upstream sync, writing effective pull request descriptions and commit messages, following project conventions, and communicating professionally with maintainers. Enforces three rules that are never skipped: work on a feature branch, test thoroughly with evidence before opening the PR, and keep each PR focused on a single feature. Documents 16 common mistakes, especially leaking personal development artifacts (SESSION.md, NOTES.md, TODO.md, planning docs, debug screenshots, temporary test files) and mixing in unrelated changes. Includes validation scripts, PR and commit message templates, and pre-submission checklists. Use when creating a pull request for a public repository or contributing to a community project.
OpenAPI contract validation between frontend and backend. Covers type sync, endpoint validation, and contract-first development. USE WHEN: user asks about "OpenAPI validation", "API contract", "swagger sync", "frontend backend types", "API mismatch", "contract testing" DO NOT USE FOR: GraphQL - use `graphql-contract` skill, authentication - use `auth-flow-validation` skill
Operate Orca-managed worktrees, folder contexts, terminals, repos, automations, artifacts, skill sharing, worktree comments, and Orca's embedded browser through the `orca` CLI. Use when the user says "$orca-cli", "Orca worktree", "child worktree", "spawn codex/claude in a worktree", "read/wait/send Orca terminal", "handoff" / "handover" / "give this to another agent", "Orca browser", "orca artifacts", or "share skills". Prefer it over raw git worktree, ad hoc PTYs, or Computer Use when Orca state is involved. Use Computer Use only when a visible window needs GUI control that a CLI, filesystem, or API cannot do.
Use the public `orca` CLI to operate Orca-managed worktrees, folder contexts, terminals, repos, automations, artifacts, worktree comments, and the browser embedded inside the Orca app. Use when the user says "$orca-cli", "use orca cli", "Orca worktree", "child worktree", "cardStatus", "spawn codex/claude in a worktree", "read/wait/send Orca terminal", "terminal send", "full handoff", "handover", "give this to another agent", "another worktree", "Orca browser", "orca artifacts", "share HTML/Markdown", "public artifact link", or "control the browser inside Orca". Prefer this over raw `git worktree`, ad hoc PTYs, Playwright, or Computer Use when the task touches Orca-managed state. Use Computer Use for browser windows, webviews, or desktop UI outside Orca's embedded browser.
Reorganize the current branch's work into a clean, reviewer-friendly sequence of bisectable commits. Use when the user asks to clean up commits, split a messy WIP into logical commits, prepare a branch for review/PR, organize changes into bisectable history, or rewrite history so each commit builds and passes tests on its own. Triggers on phrases like "organize commits", "bisectable commits", "clean up history", "split this commit", "커밋 정리해줘", "커밋 쪼개줘", "리뷰어 친화적으로".
oRPC (v1.12+) typesafe API layer for TypeScript. Covers contract-first development, procedure definition, middleware, routers, context, error handling, Hono integration, TanStack Query, event iterators, plugins, file uploads, WebSocket adapter, and best practices. Use when building or modifying API procedures, data fetching hooks, server handlers, or real-time features. Triggers on tasks involving oRPC, typesafe APIs, RPC procedures, or server route handlers.
API pagination patterns. Offset-based, cursor-based, keyset pagination. Filtering, sorting, and page metadata. REST and GraphQL pagination implementations. USE WHEN: user mentions "pagination", "paginate", "cursor", "offset", "page size", "next page", "infinite scroll API", "list endpoint" DO NOT USE FOR: frontend infinite scroll UI - use frontend framework skills; database query optimization - use database skills
React SPA frontend setup skill (PocketBase integration). Vite + TanStack Router (file-based routing) + TanStack Query + Tailwind CSS / Shadcn UI + Biome. Use when creating a React frontend, SPA setup, PocketBase UI, frontend project initialization, React app, or dashboard build.
Use when the user has a coding or engineering prompt and wants it refined into a detailed, executable plan before any code is written — the planning stage of a prompt → plan → execute → debug pipeline. It decomposes the prompt from first principles (objective, end state, environment, building blocks, tools, packages), breaks the work into PR-sized tasks each tied to a component with its files, tests, and dependencies, orders them topologically, splits each into atomic subtasks, then a separate principal-engineer agent critiques the plan for alignment, coverage, sizing, and executability; it revises until the critique passes, emitting a plan.md layout and a structured tasks.json that a junior engineer or a smaller model can execute correctly. Not for executing, scaffolding, or debugging the plan (those are downstream loops), and not for research proposals or experiment plans.
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.