Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Use when user asks about design skills, UI skills, design tools, "what design skills are available", "help me design", "recommend a design skill", "UI/UX skill", "design workflow", "設計技能", "UI 技能", "設計工具推薦", or when the user needs guidance on which design skill to use for a specific task.
This skill helps an LLM generate correct AxAgent memory retrieval, context-map, and dynamic skill-loading code using @ax-llm/ax. Use when the user asks about contextMap, AxAgentContextMap, onMemoriesSearch, memoriesCatalog, recall(...), inputs.memories, onLoadedMemories, onUsedMemories, onSkillsSearch, skillsCatalog, AxAgentCatalogSkill, discover({ skills }), onLoadedSkills, onUsedSkills, preloaded skills, preloading memories at forward time, relevanceRanking hints, loaded memory/skill IDs, or carrying memories across forward() calls.
This skill helps an LLM generate correct AxGen code using @ax-llm/ax. Use when the user asks about ax(), AxGen, generators, forward(), streamingForward(), validation, assertions, streaming assertions, field processors, step hooks, self-tuning, or structured outputs. For MCP clients, transports, prompts, resources, tasks, subscriptions, or authentication use ax-mcp alongside this skill.
When you report a brain page to the user — created, edited, committed, or relayed from a subagent — a working link is part of the deliverable, in the same message. Derive the path mechanically (git ls-files --full-name), push before linking, verify the link resolves when a hosted remote exists, and degrade through a defined fallback chain when it doesn't. Inside brain pages the rule inverts: relative links preserve the link graph; absolute URLs are for chat deliverables only.
When the user corrects a factual error, immediately investigate its root cause. Don't just note the correction — trace the error to its source, fix the source, and prevent recurrence. Every factual error is either a data error (bad brain page, bad memory file, bad rendered SOUL/USER identity, bad facts row) or a hallucination (LLM confabulated from partial signals).
Analyze tooling overlap and optimization opportunities in AI-assisted development projects. Use when: (1) auditing project tooling ecosystem, (2) identifying duplicate or redundant tools, (3) finding consolidation opportunities, (4) generating tooling health reports.
Guide a consent-based conversation that helps a person discover how an AI agent could improve their day-to-day life: routines, friction, attention, decisions, relationships, learning, and small experiments. Use when someone asks for a daily check-in, wants the agent to learn how they work, says "grill me," wants a conversational journal, or asks what an AI could help with. Do not use for therapy, diagnosis, crisis support, covert monitoring, or product requirements interviews.
Use when needing to scrape documentation websites into markdown for AI context. Triggers on "scrape docs", "download documentation", "get docs for [library]", or creating local copies of online documentation. CRITICAL - always analyze sitemap first before scraping.
Orchestrate a complete PhD / research-program application package end-to-end — from reading the applicant's materials through CV, program & faculty fit, SOP, recommender strategy, professor outreach, a final package audit, and interview preparation once shortlists arrive. Use whenever the user wants to drive the whole application rather than one piece: "help me with my PhD applications", "where do I start with my application", "manage my application process", "I'm applying to PhD programs, what's the plan", "get my materials ready", or hands over a folder of CV/transcripts/notes and asks for the path to submission. This is a ROUTER that sequences existing singleton skills with stage gates; it does not write the SOP, CV, or emails itself. For a single piece, invoke that skill directly (e.g. apply-sop-writer, apply-cv-builder). For driving a research paper use flow-paper-lifecycle.
Route genuinely ambiguous or multi-stage geospatial work across specialist skills while enforcing shared CRS, validity, leakage, units, verification, and reproducibility rules. Use for requests spanning multiple stages such as acquisition, imagery, modeling, analysis, and map delivery, or for an explicit end-to-end pipeline. Never invoke for one domain merely because a parameter is unclear. Code implementation/review, backend or platform choice, and production-readiness review are direct specialist tasks. Do not add this skill as a layer around one specialist.
Use this skill for persistent PostgreSQL storage in agent workflows — spin up databases on demand, test schema migrations safely on forks, and pass databases as assets between agents using Ghost. **Trigger when user asks to:** - Create a new PostgreSQL database quickly - Manage database lifecycle in agent workflows - Fork a database to test changes safely - Run analytics or heavy queries on a fork without impacting the production database - Share a database snapshot so others can spin up their own copy - Pass a database as an input or output between agents - Create a database for agents - Need many databases that are billed only when queried - Set up the Ghost MCP server or CLI **Note:** If the Ghost MCP server is already installed, use its tools directly — this skill is for environments without the MCP or as a reference for what's possible. **Keywords:** Ghost, ghost.build, database, create database, fork database, PostgreSQL, managed Postgres, MCP, agent database, connection string
Clarify fuzzy intent and pressure-test design choices through focused dialogue before implementation. Use when the user wants to shape an idea, resolve consequential tradeoffs, or says to grill or crystallize it. Not written-artifact critique, bug diagnosis, or implementation.
Drive the grok build CLI (run `grok`) for search, research, and media creation. Use for Grok/X web search with citations; large-scale parallel research via grok headless + subagents; generating images (image_gen) and videos (image_to_video) through the logged-in grok session at zero API cost; and ffmpeg post-processing like green-screen / chroma-key removal, concatenation, and watermarking. Cross-platform (macOS/Linux/Windows). Optional xAI search/media API as a paid fallback. Triggers include grok, grok build, 生图, 生成图片, 生成视频, 绿幕抠像, 批量调研, xAI search.
End-of-session ritual. Captures decisions, lessons, gotchas, and open threads. Writes a narrative session log to ~/.origin/sessions/ and stores granular memories via Origin MCP. Previews any unconfirmed captures from the current session before closing. Invoked as `/handoff`.
Token-reduction skill for coding agents. Compresses JSON tool output, build logs, grep/ripgrep search results, source code, and conversation history the same way the Headroom proxy does — but in-context, with no binary install. Use this skill whenever an agent is about to paste a large tool result (a 1000-row JSON array, a 5000-line pytest log, a 200-result grep, a 1000-line source file) into its next message, or when the agent's conversation is approaching the context limit, or when the user complains about token cost or latency. Also use it to shape the agent's own output: skip ceremony, don't restate code already on screen, drop reasoning on routine turns. Triggers: "compress this output", "this is too long", "trim", "summarize the log", "reduce tokens", "context is full", "running out of context", "too many tokens", "make it shorter", "be terse"
Implement a GitHub issue end-to-end using the team-executor skill. Given an issue number (or URL), opens a feature branch, dispatches a planning team specialized to the project stack to decide the implementation path (unless the issue dictates the path), confirms a TL;DR with the user (or decides autonomously when asked), posts the agreed approach as an issue comment, launches a fresh execution team to build it, and opens a pull request. Use when the user provides only a GitHub issue number, an issue URL, or asks to "implement issue #N", "work on issue N", "ship issue N", "do issue N", or "build out the issue". Bare numeric input from the user should be treated as a GitHub issue reference in the current repo.
Initialize projects with agentic coding structure. Use when setting up a new project, adding AI agent support to existing project, or when user says "init", "initialize", "setup project", or "scaffold". Creates AGENTS folder, documentation templates, and _NOTES scratch space.
INVOKE THIS SKILL when asked about and/or have to implement package versions, installation, or dependency management for LangChain, LangGraph, LangSmith, or Deep Agents. Covers required packages, minimum versions, environment requirements, versioning best practices, and common community tool packages for both Python and TypeScript.
Use when the user asks to run or continue an autonomous build: turn a goal into an approved plan, dispatch implementation and review agents, and finish with one pull request. The Lead is a provider-neutral orchestrator that makes decisions but never writes code, edits files, runs commands, or reviews its own work.
Runs a local council when no external AI providers are configured. Spawns N independent Claude subagents (one per role, blind to each other) that each answer the question from a single assigned lens, then synthesizes their perspectives. Invoked by the ask command via --local, or when the user accepts the local-council offer after no providers are found. This is a same-model (Claude-only) panel, not a cross-vendor council.
Clean, compact, and reconcile durable project context and any configured issue-tracker mirror without changing established meaning. Always use this skill whenever a request cleans, audits, or checks consistency across two or more durable context sources, even when they are called product docs, glossary, decisions, specs, or agent instructions instead of PRODUCT.md, GLOSSARY.md, docs/decisions, docs/specs, AGENTS.md, or CLAUDE.md. Also use for confirmed shipped-spec retirement or a full consistency sweep between default-branch spec folders and their tracker issues. Do not use to define new product intent, record one newly settled term or decision, look up context, shape a feature, implement work, or revise an active spec's behavior against work that has shipped, which belongs to `babysit-specs`.
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).
Run a periodic hygiene ceremony over whatever persistent memory the agent can reach — vector memory stores, knowledge graphs, diaries. Dedup near-duplicates, decay stale items, split bloated ones, invalidate superseded facts, reconnect orphans, write an interval summary, and report. Use when asked to garden/tend/clean agent memory, run memory hygiene, or as a scheduled unattended task. Storage-agnostic — discovers capabilities from the tool surface, never assumes specific tool names.
Multi-perspective deep analysis. Several sub-agents each act as advisers with different thinking frameworks, analyse the same material independently, then their consensus and disagreements are cross-summarised into a structured diagnostic report. Triggered when the user says "help me analyse from multiple perspectives", "multi-dimensional analysis", "look at it from several angles", or "help me diagnose this".
Generate evaluation datasets from QA/dev specification documents. Analyzes spec sheets, discovers agent span structure from code, recommends evaluators, auto-configures variable mappings, and produces an editable plan for review. Use when the user says: NDD generate, create test suite, generate dataset from spec, create evaluation plan, generate NDD plan, or provides a QA specification document.
(NS) Ad-hoc coding worker — bug fixes, small refactors, scripts, migrations — without full SDD. Entry priority 5: use for \"just implement this\", \"quick fix\", or concrete coding without execution-handoff (also C2 under ns-code-autonomous). Do NOT use for GitLab ISSUE_URL (ns-execution-gitlab-issue), multi-day/version scope (ns-spec-driven), diagnosis-only (ns-code-investigator), or when execution-handoff.md exists. Do NOT generate requirements/tasks/handoff.
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.