Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Lay the keel of the shipyard harness in any repo — the 4-pillar shared environment (Context, Rules, Tools, Standards), seeded as two faces on day one (CLAUDE.md + GLOSSARY.md) with every other surface appearing on first use, so that every human and agent inherits the same design language and anyone can ship. Run once per repo; re-run with --check to audit drift.
Scan the agent-framework frontier over the past 30 days and compare overstack across 8 axes (frontier-gap-scan runbook) — invoke instantly, not as a cron job. Run 5 WebSearches, rate Ahead/Close/Behind, diff against the previous period, raise issues for new gaps or update existing ones, and update the report. Triggers: 'frontier scan', 'quét đối thủ', 'scout tuần', 'chúng ta thua gì', '/frontier-scan'.
Tears down server mode for html-skills for THIS Claude Code session — kills the receiver, stops the `Monitor` armed by `html-skills-listen`, and cleans up its temp files. Parallel sessions are unaffected. Invoke when the user is done with html-skills artifacts ("I'm done with html-skills", "stop the listener", "end html-skills session"), or at session end if the user mentions cleanup. Safe to invoke when nothing is active — it reports inactive and exits
Shipyard's governed delivery pipeline — converge the mission, synthesize a durable spec, decompose vertical-slice tickets with blocking edges, run the frontier in parallel via team, close with verification, and report with a full decision log. Two entry gates — the yard gate (drydock audit) and the fog gate (an effort whose destination is unclear is routed to /ask-navigator before this pipeline starts). Humans own the checkpoints where there is no unique answer or the error cost is severe; agents continuously run everything repeatable and acceptable-by-evidence
The composed executor that lands a subagent worktree branch onto an integration branch with preflight + recorded recovery point. For a shared integration branch route through `serialize_merge` (the single-writer lease that composes this action); invoke `merge_orchestrate` directly only for a non-integration merge or when you already hold the lease. Triggers: operator (or `next_actions`) surfaces the merge verb via Exarchos MCP. Local git operation — NOT remote PR merging (that is `merge_pr`).
Convert a MuleSoft Agent Network V1 project (`schemaVersion: 1.0.0`) into a V2 project (`agentNetwork: 2.0.0`). Each V1 broker becomes a V2 broker with one orchestrator node — no graph splitting, no prompt rewriting. Use when the user asks to upgrade, convert, migrate, translate, or port an Agent Network / Agent Broker / Daisy planner from V1 to V2, or whenever the working folder contains a V1 `agent-network.yaml`.
Ghi một SESSION EPISODE có cấu trúc (tầng nhớ episodic) vào memory store để phiên sau truy hồi được 'phiên trước làm gì' — theo NGỮ NGHĨA, không chỉ theo [[wikilink]]. Bọc mem-rank.py (engine). On-demand, KHÔNG auto-hook (ADR-004). Trigger: 'ghi episode', 'record episode', 'chốt phiên này vào nhớ', 'lưu lại phiên trước làm gì', 'session episode', '/record-episode'.
Route multi-step marketing, product, or business challenges to the right sequence of ClawFu skills. Recommends which frameworks to combine, in what order, with handoff outputs between steps. Use when: planning a product launch end-to-end, combining positioning + offers + launch skills, running a customer-validation sprint, creating a content strategy pipeline, deciding which skills to chain for a complex project, or orchestrating a sales-enablement playbook.
Reference for using Plannotator (its `plannotator` tool when available, otherwise the CLI): plan review, code review, annotating files, URLs, folders, and running local apps, annotating the last assistant message, browsing archived plan decisions, exporting or sharing Guided Reviews, and the Plannotator Inbox (messages and questions the person answers later, without blocking you). Invoke when asked to use Plannotator for anything not covered by a more specific plannotator-* skill, and before leaving the person a decision, question, or finished work while they are away from the chat (if another agent started you, report to it instead)
Writes, refactors, and evaluates prompts for LLMs — generating optimized prompt templates, structured output schemas, evaluation rubrics, and test suites. Use when designing prompts for new LLM applications, refactoring existing prompts for better accuracy or token efficiency, implementing chain-of-thought or few-shot learning, creating system prompts with personas and guardrails, building JSON/function-calling schemas, or developing prompt evaluation frameworks to measure and improve model performance.
Cross-tool golden rules for an autonomous Scrapfly web agent connected to the Scrapfly MCP server (web_scrape, screenshot, check_if_blocked, the Cloud Browser lifecycle, snapshot/click/fill/type/press/scroll/select/drag, and WebMCP). Use when an LLM agent has access to the Scrapfly MCP toolset and needs to choose between stateless scraping and a stateful Cloud Browser session, decide when to call browser_unblock, or coordinate multi-tool flows
Ghi và kiểm provenance (nguồn + sha256 checksum) cho skill — dùng khi 'cài skill từ ngoài', 'skill này từ đâu', 'audit nguồn skill', 'checksum skill', 'supply-chain skill', phát hiện skill bị sửa lén; bổ trợ orca-sec-scans. Chạy fdk/tools/skill-provenance.py. /skill-provenance
Use when auditing an agent codebase against the 12-Factor Agents methodology, reviewing LLM-powered system architecture, or assessing agentic app compliance. Triggers on "analyze agent architecture", "12-factor audit", "how compliant is this agent", or "evaluate this LLM app". Also applies when comparing frameworks or planning agent improvements. Not for quick checklists — this performs deep per-factor codebase analysis with file-level evidence.
Route product-growth requests to one Gingiris specialist skill. Use when broad or explicit requests about launch, Product Hunt, GitHub/open-source growth, SEO/GEO, B2B/PLG, ASO, interviews, competitors, KOL, UGC, community, content, backlinks, or README conversion, including 出海增长、产品发布、找增长渠道、竞品分析、 用户访谈、社区运营. Ask short multiple-choice questions only when the route is unclear; then load and run one specialist rather than the full collection.
Ultra-compressed response mode. Cuts token usage by dropping articles, filler, pleasantries, and hedging. Uses symbols for relationships. Technical terms and code blocks remain exact and uncompressed. Use when user says “save tokens”, “RTU mode”, or “compressed response mode”.
Use this skill when the user wants to “create a skill,” “add a skill to a plugin,” “write a new skill,” “improve a skill description,” or “organize skill content,” or needs guidance on skill structure, progressive disclosure, or skill development best practices for Claude Code plugins.
Turn user-provided requests, drafts, or prompt text into clearer, more executable prompts without performing the underlying task. Use when the user explicitly wants to improve, rewrite, optimize, tighten, clarify, structure, or adapt a prompt, when they have an idea but do not know how to phrase it, or when the prompt is missing audience, constraints, format, execution context, or success criteria. Also use for system prompts, meta-prompting, prompt frameworks, Claude Code prompts, and batch or API prompt design. Do not use when the user mainly wants the task executed, researched, or implemented rather than the prompt improved.
Vercel AI SDK v6 development, for projects already on ai@6. Use when building or maintaining AI agents, chatbots, tool integrations, streaming apps, or structured output in a v6 codebase. New projects and ai@7 code use ai-sdk-7; an unknown version goes through ai-sdk. Covers ToolLoopAgent, useChat, generateText, streamText, tool approval, provider tools, MCP integration, and Output patterns.
Helps users discover, search and install agent skills from the marketplace. Use when the user wants to find a skill, discover new capabilities, find specific tools (especially sandboxes) or install a skill. This skill should be used when the user is looking for functionality that might exist as an installable skill. Also supports progressive disclosure to recommend sandbox-related skills when relevant. 用于帮助用户在市场中搜索、发现和安装 Agent 插件（Skills）。 当用户想要查找功能、安装特定工具或寻找沙箱（sandbox）时使用。 Triggers: "查找插件","搜索技能","搜索 X 技能","安装技能","安装 X 技能","如何做 X","沙箱","安全沙箱","find a skill for X", "how do i do X", "search for skill", "install skill", "I need a skill for X", "sandbox", "aio sandbox"
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.
Run structured multi-agent debates using argue CLI for cross-examined, high-confidence answers. Use when facing strategic decisions, ambiguous trade-offs, architecture debates, or questions where multiple perspectives improve the answer. Triggers on: argue, debate, cross-examine, second opinion, multi-agent, 'Should we X or Y?' with real stakes, consensus-building, risk analysis, or confirmation-bias mitigation.
Nuwa creates people: enter a person's name or topic, or even just a vague need, to automatically conduct in-depth research → distill a thinking framework → generate a runnable persona Skill. Two entry paths: (1) a specific name → distill directly; (2) a vague need → diagnose and recommend → then distill. Triggers: "create a skill", "distill XX", "Nuwa", "create a person", "XX's way of thinking", "make an XX perspective", "update XX's skill". Vague needs also trigger it: "I want to improve my decision-making", "Is there a way of thinking that could help me...", "I need a thinking consultant".
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.