Skills
97,501 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Huawei Cloud Flexus X Instance Deployment - DeepSeek Harness (dsh). ⚠️ Deployment involves actual charges, must confirm region/flavor/billing mode with user before execution. Triggers include: "部署DeepSeek Harness", "部署dsh", "部署DeepSeek智能体", "创建AI智能体服务器", "Deploy DeepSeek Harness", "Deploy dsh", "Create AI Agent Server", "Deploy AI Harness", "DeepSeek Harness", "dsh", "Flexus X", "ECS部署", "AI智能体".
Design and create Busabase-backed App-in-Skill packages with a canonical app project, Busabase-native setup and product onboarding, review/execution workflows, a standardized responsive UI, and AirApp-first delivery. Use when a user wants a Busabase research desk, review queue, planner, action console, operating dashboard, control panel, collaboration workspace, onboarding/readiness flow, or an existing Kelly App-based skill updated. Every generated skill contains a complete app/ project, deploys that source to Busabase AirApp by default, runs pnpm dev only when local preview is explicitly requested, follows the Kelly desktop and phone UI contract, delegates AirApp runtime, SDK, security, scaffolding, and deployment constraints to busabase-app-creator, and delegates repository-level conformance testing to kelly-app-skill-creator-tests.
Operate KWeaver knowledge networks and Decision Agents — build knowledge networks, query schemas/instances, semantic search, execute Actions, Agent CRUD and conversations, Trace data analysis. Operate Dataflow document processes — list flows, trigger runs, query run history, view step logs. Operate the Skill management module — register Skills, market lookup, progressive reading, download and install. Operate Toolbox / Tool — create toolboxes, upload OpenAPI tools, publish and start/stop. Operate the Vega observability platform — query Catalogs/resource/connector types, health checks. Operate the model factory — LLM/small model CRUD (mf-model-manager), OpenAI-compatible chat and small model invocation (mf-model-api), adapter registration. Automatically used when the user mentions intents such as "knowledge network", "knowledge graph", "query object class", "execute Action", "which Agents exist", "create Agent", "chat with Agent", "list all Agent templates", "list Agents I created", "list Agents in my private space", "dataflow", "data flow", "process orchestration", "process run records", "process logs", "trigger dataflow", "view dataflow run history", "Skill", "skill package", "register Skill", "install Skill", "read SKILL.md", "toolbox", "upload tool", "register tool", "OpenAPI tool", "enable tool", "publish toolbox", "data source", "data view", "atomic view", "Catalog", "Vega", "health check", "inspection", "trace", "evidence chain", "data flow tracking", "data source", "how the data was obtained", "BKN metric", "metric management", "metric query", "metric dry-run", "model factory", "small model", "embedding", "rerank", "mf-model", "model chat".
Migrates deprecated Letta Filesystem folders/files to MemFS using markdown document corpora, chunking, local lexical search, and QMD semantic search via the memfs-search skill. Use when replacing folders.files.upload, working with PDFs or document QA, or emulating open_file, grep_file, and search_file behavior.
Use the LiveKit CLI to manage LiveKit Cloud projects, deploy agents, generate tokens, and configure telephony. Use this skill when working with the `lk` command-line tool for project setup, agent deployment, phone number management, or SIP configuration.
(NS) Generate or refresh project AGENTS.md plus CLAUDE.md (Rules boot + AGENTS.md + .claude/agents) after harness install. Use for "write AGENTS.md", bootstrap agent entry point, replace harness stub, or configure Cursor/Claude project rules. Do NOT copy the harness template verbatim. Do NOT use for architecture-rules, reverse business specs, or brownfield maps.
(NS) Resolve ambiguities and gaps in a vague scope before requirements. On brownfield, runs bootstrap when brownfield-map.md is missing; if map exists, human gate to refresh or keep. Use for incomplete scopes ("dashboard", "notifications", "integration with X") or before requirements when terms are unclear. Do NOT generate requirements.md — enriched context only.
Comprehensive guide for integrating Obsidian vaults as AI-powered second brains with Claude Code. Covers MCP integration, vault manifests, self-evolving patterns, auto-linking, and knowledge graph automation. Use when setting up Claude Code + Obsidian workflows, implementing bidirectional sync, creating CLAUDE.md manifests, or building self-healing knowledge systems.
OpenClaw operations assistant. Use when the user mentions OpenClaw, 小龙虾, gateway, channel connections, failed message sending, unreachable services, log troubleshooting, or channel or Agent management. Prioritize status checks and fault triage; for high-impact operations such as restarting, repair, updates or configuration changes, explain them to the user before executing.
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.
Reusable approval-loop pattern for fanning out lead personalization across parallel Claude Code Task sub-agents. Shows the user 1 sample personalization, collects feedback, runs 10 more, approves, runs 10 more — stops when 2 consecutive rounds have zero edits, then scales to the full list. ALWAYS uses Claude Code Task tool sub-agents — never an external Anthropic/OpenAI API key. Use when any skill needs per-lead custom variables (situation lines, value lines, CTAs).
The main entry point for rayskills: use when unsure which ray-* skill to use, given only a real task, or when taking content from idea through to the draft box or a spoken script. Helps determine the next step or execute a confirmed production chain. “Push” or “run the full process” does not mean authorization to publish. Trigger: /ray
Execute approved implementation plans phase by phase with automated and manual verification. Use when the user explicitly says "implement the plan", "execute the plan", or "start implementing" and has a plan file ready. Do not use for ad-hoc coding tasks without a plan.
Browses and install third-party marketing, research, and creative skills. Use when: 'install skills', 'available skills', 'third party skills', 'marketing skills', 'opc skills', 'what skills are available', 'list skills', 'browse skills', 'show me skills'. Do NOT use for routing a specific sales objective to the right skill (use /sales-do) or requesting a brand-new skill that doesn't exist yet (use /sales-request-skill).
Organize persistent coding-agent guidance by scope and source of truth. Use when creating, converting, reviewing, or reorganizing AGENTS.md, CLAUDE.md, custom-agent wrappers, or repo-local skills, or when preventing duplicated guidance across tools.
Coordinate team task execution on OpenAnt. Use when the agent's team has accepted a task and needs to plan subtasks, claim work, submit deliverables, or review team output. Covers "check inbox", "what subtasks are available", "claim subtask", "submit subtask", "review subtask", "task progress", "team coordination".
Catalog of common mistakes that break Claude Code agent and skill triggering, plus marketplace validation rules. PROACTIVELY activate for: (1) plugins installed but never trigger, (2) agents failing to route, (3) skills not appearing in discovery, (4) pre-release triggering audit, (5) migrating deprecated agent true flag, (6) zero-frontmatter SKILL.md files, (7) boilerplate in YAML descriptions, (8) abstract descriptions, (9) missing example blocks on fat agents above 2,500 words, (10) missing PROACTIVELY or Provides sections, (11) running validate_plugins.py, (12) auditing a plugin without contradicting earlier intentional refactors. Provides: anti-patterns, fixes, before/after examples, audit checklist, and validator rule index.
Write docs/<feature>/HANDOFF.md — the briefing a fresh agent session needs to pick up unfinished work with none of the conversation, covering where the branch and checkout are, what is pushed, what to merge first, what was built and which documents are authoritative, where to resume, and the traps that cost this session time. Use when an implementing run (execute-plan, execute-roadmap, build-feature) stops with work still owed, when context is about to run out, or when the user invokes /write-handoff <feature> or asks to hand work to another agent or session. For handing finished work to a person to test and accept, use write-acceptance instead.
Brand and product foundation supervisor — coordinates routing and collaboration across market discovery, competitors, brand, product, and launch. Triggers: 品牌基建, foundation, 从零开始, from scratch, 品牌规划, 产品上市, brand building, market validation, go-to-market.
Build a minimal AG2 `Agent` end to end — pick a model provider, set a prompt, call `agent.ask()`, then continue the conversation with `reply.ask()` (multi-turn). Use when the user is starting a new AG2 project, has no working `Agent` yet, or needs the multi-turn chaining pattern. Covers `OpenAIConfig`, `AnthropicConfig`, `GeminiConfig`, `OllamaConfig` etc., and env-var fallback for API keys.
Use this skill whenever you need to automate or control Google Chrome on Linux using the user's existing authenticated session on any site — particularly sites that block automated or headless sign-in (MFA, CAPTCHA, OAuth, bot detection). Trigger any time the user asks you to "use Chrome", "open the browser", "automate a website", "control the browser", "use agent-browser", or wants to do something on a site they're already logged into. Also trigger when the user mentions CDP, remote debugging, or connecting to a running Chrome instance on Linux. This skill is essential whenever agent-browser needs an authenticated session — don't skip it, since missing the profile-copy step will drop you on a sign-in page you can't get past.
Run a structured multi-agent debate by spawning a panel of expert agents on any question, with convergence-aware iteration and typed synthesis output via the `agent-council` CLI. Use when a decision has genuine tradeoffs, high stakes, or hidden assumptions worth adversarial collaboration, or when confidence diagnostics matter more than a single recommendation. Compatible with any AI agent harness that supports agentskills.io skills (Claude Code, Cursor, Hermes Agent, OpenHands, etc.). Do not use for simple factual lookups, tasks with a clear correct answer, routine single-perspective work, or code execution and tool orchestration beyond debate.
Generate project-level AGENTS.md guides that capture conventions, workflows, and required follow-up tasks. Use when a repository needs clear agent onboarding covering structure, tooling, testing, task flow, README expectations, and conventional commit summaries.
Use when automating a real, logged-in Chrome WITHOUT disturbing the user's open tabs — e.g. authenticated sessions like ChatGPT, LinkedIn, or any site needing a persistent login. Launches or reuses a dedicated debug Chrome instance per named profile that coexists with the user's main browser; it never quits or kills any browser (unlike real-browser, which closes Chrome Beta's tabs). Trigger on "automate this site without closing my tabs", "use my logged-in session", "open ChatGPT/LinkedIn in automation", or when a browser task must preserve the user's existing windows. macOS, requires the agent-browser CLI.
Turn a plain-language request into a working, tested Agenta agent through the Agenta API. Use when the user wants to build, create, schedule, or test an Agenta agent from their coding agent (Claude Code, Codex, Cursor). Writes one config and runs a few bundled shell scripts. Triggers on "build an Agenta agent", "create an agent on Agenta", "schedule an Agenta agent", or wiring an agent to a tool like GitHub or Slack via Agenta.
Use when an agent needs to operate the user's real Chrome session — listing tabs, snapshotting the page into actionable @refs, clicking, filling, typing into rich editors, pressing keys, evaluating JS, capturing screenshots, and reading console/network buffers. All actions go through CDP and run on backgrounded tabs without stealing focus.
Batch-sync locally developed Skills to both ClawHub and Tencent SkillHub. Supports smart .gitignore filtering, whitelist control, incremental sync, single-skill sync, and split publishing across the two platforms. Use this skill when the user needs to publish local skills to ClawHub/SkillHub, batch-sync skills, or check publishing status.
Deep interview skill for shaping a Clay Mate's identity through conversation. This skill runs automatically when a new Mate is created through the creation wizard, or when a user wants to reshape an existing Mate. The Mate itself conducts the interview, asking questions to understand the user and building its own CLAUDE.md identity file. Use this skill whenever a Mate DM session starts for the first time, when the user says things like "let's redo your personality", "I want to change how you work", "reshape", "re-interview", or "start over". Also triggers on /mate-interview or when Clay's Mate creation flow injects seed data into a new DM session.
Create a Bolna Voice AI agent with the current v2 API, including system prompt, welcome message, LLM, voice, transcriber, telephony input and output, latency tuning, guardrails, knowledge bases, and function tools. Use when the user wants to build, deploy, or clone a Bolna voice agent.
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.