Skills
97,501 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
This skill should be used when users want to route LLM requests to different AI providers (OpenAI, Grok/xAI, Groq, DeepSeek, OpenRouter) using SwiftOpenAI-CLI. Use this skill when users ask to "use grok", "ask grok", "use groq", "ask deepseek", or any similar request to query a specific LLM provider in agent mode
Use when resuming work after a pause, reboot, or context switch and the workspace uses `wiki/checkpoints/` resumable notes. Read the latest relevant checkpoint chain, orient to the most likely in-flight scope, verify current files and tools before acting, and report the safest next step. Treats a goal path as concrete context; checks live goal status over stale checkpoints.
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).
(NS) Execute a GitLab issue end-to-end — status gates, worktree isolation, atomic delivery, MR, mandatory code then judge loop. Use for external ISSUE_URL or SDD delivery unit (unit + issue_iid). External Phase 2: single-unit → coder-agent in worktree; multi-unit → ns-autonomous Engine. SDD unit mode uses ns-coder run-implementation. Needs mcp-gitlab-usage + ns-reviewer + ns-judge.
Run Comet's Opik — open-source LLM observability, evaluation, and optimization — from one routing-first skill: install the Python/TypeScript SDK, stand up a server (Comet.com cloud, Docker Compose via `./opik.sh`, or Kubernetes/Helm), wire tracing through `@opik.track` or one of 50+ framework integrations (OpenAI, Anthropic, LangChain, LangGraph, LlamaIndex, CrewAI, DSPy, Ollama, Bedrock, Vercel AI SDK, …), score outputs with LLM-as-a-judge metrics (Hallucination, Moderation, Answer Relevance, Context Precision), and run Datasets/Experiments evaluations including PyTest CI gates. Use when the user wants LLM tracing, prompt evaluation, production LLM monitoring, agent optimization, or guardrails with Opik. Triggers on: opik, comet opik, opik configure, opik.sh, llm observability, llm tracing, llm as a judge, hallucination metric, prompt evaluation, opik dashboard, opik guardrails, agent optimizer.
Run 3+ genuinely independent build tasks concurrently on Claude Code's real parallel agent runtime—throughput without file conflicts or fake orchestration. Proves independence (disjoint write scope, no dependency edges) before launching, matches each lane to the best specialized agent, writes a complete self-contained brief per fresh agent, launches them all in one message with background execution and worktree isolation, then conflict-checks and reviews each finished lane before treating it as merged. Refuses to parallelize fewer than 3 lanes or lanes that share write scope or dependencies, never assumes fresh agents inherit context, and reports partial success per lane rather than a blurred merged summary. Use when an approved plan or the user gives 3+ independent build lanes.
Install the ToolUniverse Claude Code plugin in one step — provides MCP server with 1000+ scientific tools, 120+ research skills, slash commands, hooks, and the research agent. Use for first-time plugin install, troubleshooting plugin not loading, verifying MCP server connection, listing API key requirements, or configuring auto-update
Write and optimize prompts for AI-generated outcomes across text and image models. Use when crafting prompts for LLMs (Claude, GPT, Gemini), image generators (Midjourney, DALL-E, Stable Diffusion, Imagen, Flux), or video generators (Veo, Runway). Covers prompt structure, style keywords, negative prompts, chain-of-thought, few-shot examples, iterative refinement, and domain-specific patterns for marketing, code, and creative writing.
Create or update a repo's DESIGN.md — a living design-language contract for AI agents and humans — plus a portable frontend-design skill. Classifies the repo first (document an existing design language, merge with existing design docs, or propose a new direction with user approval) so it never overwrites a project's current design. Use when setting up design docs for a frontend project. Triggers on "add a DESIGN.md", "document our design language", "set up design docs", "create a design system doc", "bootstrap design".
Challenges AI-generated plans, code, and designs via pre-mortem, inversion, and Socratic questioning to surface blind spots and failure modes. Triggers on: "challenge this", "devils advocate", "stress test this plan", "poke holes in this", "what am I missing".
Drives a gm walk (nodes, PRD rows, pool and hop subagents) for coding, refactoring, debugging and engineering work. Use for a gm walk, a PRD row, or the pool protocol. Does not cover one-off questions, read-only lookups, code review (code-review), simplification (simplify), or launching the app (run)
Use when pausing, shutting down, handing off, or context-switching active work and future sessions need a compact resumable checkpoint derived from the current session context. Writes a small checkpoint note under wiki/checkpoints/ and then optionally records the user's intended return step. Not for durable learnings capture, wiki correction, or source ingestion.
Use when an approved workspace spec needs an execution-ready implementation plan with ordered, verifiable steps. Specs are mandatory: this skill consumes design decisions, verifies that the spec still matches the codebase, and writes the repo-native plans/ artifacts. Inherits optional goal provenance; see loam::setting-goals.
Use before making non-trivial assumptions — before touching a subsystem you have not this session, before running or changing a command pipeline, when code and docs conflict, when asked "what is the right way here?", when debugging a recurring failure, or when about to answer with "probably/usually/likely".
Search, evaluate, and recommend agent skills from the skills.sh directory. Use when: (1) user asks to find/search a skill for X, (2) "есть скилл для...", (3) "skill for zoom/notion/stripe/...", (4) "поищи на skills.sh", (5) "нужен плагин/расширение/скилл для...", (6) "как подключить X к Claude Code", (7) any mention of skills.sh, (8) "найди скилл", (9) "find skill", (10) user shares a skills.sh link for review.
Token-Oriented Object Notation (TOON) format expert for 30-60% token savings on structured data. Auto-applies to arrays with 5+ items, tables, logs, API responses, database results. Supports tabular, inline, and expanded formats with comma/tab/pipe delimiters. Triggers on large JSON, data optimization, token reduction, structured data, arrays, tables, logs, metrics, TOON.
Resume work on the current task. Loads the workflow Phase Index, figures out which phase/step to pick up at, then pulls the step-level detail via get_context.py --mode phase. Use when coming back to an in-progress task and you need to know what to do next.
Answer questions about the AI SDK and help build AI-powered features. Use when developers ask about Vercel AI SDK, generateText, streamText, ToolLoopAgent, useChat, providers, tools, structured output, embeddings, streaming, or adding AI to an app. First identify the installed major version and route version-specific work: use ai-sdk-7 for AI SDK 7 features/migrations such as WorkflowAgent, HarnessAgent, reasoning, runtime/tools context, toolApproval, telemetry, realtime, or v6-to-v7 upgrades; use ai-sdk-6 for v6 code.
Tool use, Agent Skills, and MCP for the Claude API (platform.claude.com). Tool definitions (JSON Schema), tool_use/tool_result, server tools (code execution, web search, web fetch, memory tool, tool search), client toolset (computer use, browser use), Agent Skills (SKILL.md), Skills API GA (/v1/skills), MCP connector, mcp_servers, mcp_tool_listing, MCP tunnels, and Tunnels API.
Show AI coding usage insights: cost, tokens, session activity, and breakdowns by member, model, agent, project, product, or prompt quality. Also finds who on a Team (CawPlan product line) hasn't submitted their daily coding report, rolls up a Team's cost, flags high-cost/low-diff sessions, and flags a member's cost growth. Use when: the user asks about coding costs, AI usage stats, team session activity, model spend, agent usage, prompt quality, productivity metrics, which team members haven't submitted/committed a coding report, a Team's total coding cost, sessions with disproportionately high cost vs code changed, or whether someone's coding cost has grown notably. NOT for: submitting reports, creating tickets, product health metrics, or release tracking.
Harden a plan with independent Claude/Codex review, then optionally build and cross-inspect it. Start in either Claude Code or Codex: the host plans and the other provider reviews. Use for claudex this plan, claudex-loop, or the legacy crucible trigger; not for trivial edits.
Call the Coze intelligent agent API for conversations, workflow execution, and other operations. Use when users need to integrate a Coze agent, call the Coze API, or develop Coze-related applications. Supports streaming and non-streaming conversations and workflow calls.
Gingiris growth toolkit main entry point. Call it when the user asks growth questions such as “how to drive growth,” “how to launch on PH,” “SEO traffic dropped,” “write a blog post,” or “analyze competitors.” Automatically routes to the corresponding sub-skill based on the question type (gr-seo-patrol / gr-blog-post / gr-ph-launch, etc.).
Build, evaluate, and train AI agents on RL environments with HUD. Use whenever someone wants to create an RL environment, benchmark, eval, or training task — for a coding, computer-use, browser, or robotics agent — or run and grade tasks across any model (Claude, OpenAI, Gemini, or open/self-hosted models). Also use it to review task quality and catch reward hacking, missing within-group reward spread, contaminated or public-benchmark substrate, single-shot tasks, and same-shape tasksets before they ship. Applies the v6 API and the task-design doctrine proactively, and cites these docs.
Use LeanCTX Context Gateway tools to select task-relevant code context, apply supported controls, reuse local knowledge, and inspect recorded evidence. Use when lean-ctx MCP tools (ctx_*) are available or file, search, or command output needs a focused view
(NS) Produce a codebase-grounded implementation brief for one GitLab issue or pasted scope. Blocking questions stay in chat (recommended answer, reply by number) and never go to GitLab. Use for enrich/expand requirements, grill-me, implementation brief, preview-then-post, ISSUE_URL before coding. Modes: Issue, Chat. Do NOT replace `/ns-spec-driven` Specify or version Clarify. Do NOT: PM intake, PRD, issue execution (`ns-execution-gitlab-issue`), start_execution_planning, code review (`ns-reviewer`).
Use when you discover or confirm a durable fact, constraint, invariant, convention, sharp edge, footgun, or reusable workflow that a future agent would otherwise rediscover — capture it before it is lost to conversation context. Triggers at phase/slice wrap-up too.
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.