Skills
97,501 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Show evidence-linked architecture, constraints and change scope. Default to on-demand for explicit Birdview or map requests. Also activate before coding when the project explicitly configures Birdview auto mode. Ordinary coding does not activate Birdview without that opt-in.
Manages OrchestKit learning system including feedback status, usage pattern tracking, and privacy/analytics consent. Supports pause/resume learning, data export, privacy policy display, and bug reporting. Tracks learned patterns and agent performance metrics. Use when reviewing learned patterns, pausing learning, or managing data consent.
Capture a correction or lesson as a persistent learning rule with category, mistake, and correction. Stores, categorises, and retrieves rules for future sessions. Use after mistakes or when the user says "remember this", "don't forget", "note this", or "learn from this".
START HERE after installing Brad Frost's skills — run this once to get everything set up. Finds which of the skills you installed, runs the setup each one needs (the limits interview), and tells you how to use the ones that need no setup. Use when the user says "setup brad frost skills", runs /setup-brad-frost-skills, has just installed this collection, or asks how to get started with these skills.
Provides comprehensive guidance for Spring AI including AI model integration, prompt templates, vector stores, and AI applications. Use when the user asks about Spring AI, needs to integrate AI models, implement RAG applications, or work with AI services in Spring.
Build AI-powered voice agents using Twilio ConversationRelay. Handles real-time speech recognition (ASR), text-to-speech (TTS), and bidirectional audio streaming via WebSocket. Covers TwiML setup, WebSocket message types, LLM integration, streaming responses, and voice provider configuration. Use this skill to build voice bots, IVR replacements, or real-time AI voice assistants on Twilio calls.
Apply Andrej Karpathy-inspired coding-agent guidelines in Codex. Use when writing, reviewing, debugging, or refactoring code to surface assumptions, avoid overengineering, keep edits surgical, and define verifiable success criteria.
CamoFox CLI — 50 commands for anti-detection browser automation from the terminal. Use when the user needs CLI reference for camofox commands, terminal-based browser control, or a quick lookup of command syntax. Triggers include "camofox command", "CLI reference", "terminal browser", or any request for camofox command-line usage.
End-to-end orchestration for non-trivial software feature development. Use this skill whenever the user asks to implement a PR-sized feature, break down a plan, have subagents review a plan, run a plan-review-development-acceptance loop, coordinate multiple review perspectives, produce an acceptance report, or generate an HTML PR summary. Prefer this skill for multi-step code changes even if the user only says "build this feature" and the task is not a tiny one-file edit.
Tour — Claude tự diễn cho user xem harness chặn mình theo thời gian thực trên project thật. Hai độ dài — "short" = 3 cảnh (R1·R2·R3), "full" = 10 (R1–R10). Tự phát hiện PoC vendor-neutral (cách CHÍNH) hay harness production. Tự dọn sạch demo. Trigger: "harness tour", "xem harness làm gì", "/harness-tour", "test short", "test full".
Final retrospective and self-assessment for participants of Claude Code Lab. Runs four sequential interactive parts — progress audit, best prompt, monthly plan, and feedback — using AskUserQuestion. Triggers on “/lab-retro,” “lab retrospective,” “claude code lab final,” or after completing the 6-week Claude Code Lab cohort
How to turn a working session into a MapLibre agent skill — you researched something MapLibre's docs did not cover, got it wrong before you got it right, shipped it, and the session is about to end. Covers what to capture while the context is still live, how to separate the durable claim from the project it came from, and where to hand it off. Use at the end of a MapLibre task that took real research, or when asked to capture one as a skill.
Determine the stage of the research task and route it to a primary workflow. Use when the user asks for “getting started,” “starting a research workflow,” “which workflow should this research task use,” “help me choose a research skill,” or requests the rw-research-router workflow. Runs without a private local workspace or preset research lab; use user-provided materials and bundled public-source methods.
Audits a Claude Code setup for token waste and reports only, changing no file or setting. Measures each cost driver with shell and file tools instead of estimating, covering CLAUDE.md sizes and @imports, MCP server and tool counts, whether tool deferral is active, a proxy that silently disables it, model and effort and mid-session switches, output-reducing hooks, per-agent model inheritance, cron and loop intervals versus the measured cache TTL, and the newest session log's token split. Emits one severity-ranked table plus the single highest-leverage fix. Use when the user asks to audit token or context waste, cut token cost, explain a low cache-hit rate, or diagnose context bloat. Triggers include "token waste", "token audit", "context bloat", "cache hit rate", "why is my context so big", "cut token cost".
Persist project decisions in reviewable `.agent-memory/` markdown. Use when asked to "remember a project decision", "save session learnings", "initialize .agent-memory", "manage project memory", "migrate memory", "build memory from docs", "sync memory", "run memory maintenance", "check memory status", or "onboard a new developer"; when `.agent-memory/` or `AGENTS.md` memory pointers are in play; when sharing project context, conventions, or session handoffs with another developer or reviewing project memory in a pull request; or when context "forgets between sessions". Sync reconciles files already in the checkout. Not for vendor auto-memory consolidation, personal agent rulebooks, memory leaks, RAM usage, or human memory
Install, upgrade, recommend, or combine the AlignFirst CLI, skills, docmap, and workspace in a consumer repository, or through a companion directory that leaves the repository untouched; set up AlignDev (`aligndev`), either as a repository and a Linux deployment for OpenClaw, or as a global skill that makes a coding agent the AlignFirst assistant.
Create an AI persona from scratch. Interactive flow: enter a name → provide source material (URL/file/pasted text) → automatic distillation → output a ready-to-use skill directory. Activate when the user says /create-soul, “创建人物,” “蒸馏 [某人],” or “create a persona for [name]”
Check the template’s “standard” pattern—whether it’s complete, whether there’s local drift, and whether /sync-template from remote is needed. Compare local ↔ remote ↔ disk version.json (0 token, fail-open). Triggers: “health check”, “check pattern”, “cần sync chưa”, “/health-check”.
Use when the user wants to schedule, automate, trigger, or be reminded of anything — "remind me…", "every morning…", "in two hours…", "when a new Gmail arrives", "when this record changes", "post a weekly summary". Decides between a reminder, an automation, and a skill, and builds the correct payloads.
Distill anyone’s way of thinking—their mental models, decision-making heuristics, and expressive DNA—to generate a runnable OpenClaw Perspective Skill. Fully automated pipeline: six parallel research tracks → dual-track distillation → source verification → adversarial testing → output. Trigger phrases: “distill XX,” “create a person,” “Nuwa,” “XX’s way of thinking,” “make a skill from XX’s perspective,” “perspective skill.” Vague requests also trigger it: “I want to improve the quality of my decisions,” “Is there a way of thinking that could help me…,” “I need a thinking consultant.”
End-to-end video production orchestration for the content-creation workspace. Use when the user asks to make, recreate, package, render, quality-check, or archive a video; says 制作待制作队列, 按交接稿制作, or 制作队列视频; points to a handoff project folder under 01-内容生产/视频工作台/待制作/ (with contract file 交接稿.md inside) or an image-card handoff under 01-内容生产/图文工作台/待制作/ (whose frontmatter is the binding production contract); when a request could involve HyperFrames, Remotion, captions, TTS, downloading/transcription, website/product videos, motion graphics, or reference replication; or when the user needs production planning before editing.
Generate a structured handoff document capturing current progress, open tasks, key decisions, and context needed to resume work. Use when ending a session, saying "continue later", "save progress", "session summary", or "pick up where I left off".
Audit, score, or improve an existing skill using repository checks, free Tessl lint, optional credit-aware Tessl review, metadata discovery, progressive disclosure, eval coverage, and repo conventions. Use for skill quality, activation failures, package comparisons, or publish readiness. Do not use for general code review, application verification, or unrelated documentation cleanup.
Survey Director — Responsible for topic planning, outline design, task assignment, and final review for frontier AI/ML survey papers. Activate when the user requests help writing a survey. Coordinates 5 Agents through the entire process, from topic selection to final draft.
Breaks requirements down into a structured task list and generates a task management system for long-running Agents (based on Anthropic Effective harnesses methodology). Automatically triggers when users need to manage multi-session development tasks, track feature completion, or request “break down tasks,” “task management,” or “project planning.”
Query ChatGPT/Codex reset announcements and the remaining quota and spare Full resets of multiple Pro accounts; reuse Google sign-in to verify and restore web accounts one by one, converting to Beijing time. Distinguish Tibo's official announcements, unannounced platform silent resets, banked resets and account-level periodic resets. Use when the user asks "when does it reset", "when will the quota be restored", "what time is the next everyone reset", "has the banked reset arrived", "what did Tibo say", "usage limit when reset", or says "the quota suddenly went back to 100%", "it seems to/certainly has reset again", or asks "are all the accounts used up", "which one is still full". Must cross-verify using live product status, independent user measurements and announcements; it is forbidden to deny a reset that has already happened just because Tibo Radar has no new entry, and Pacific Time must be converted to Beijing time on the spot.
Agent-first web search and retrieval for live web search, news search, docs lookup, code lookup, company research, site crawl, site map, and structured evidence packs. Includes a real no-key baseline plus optional Tavily, Exa, Querit, Serper, Brave, SerpAPI, You.com, SearXNG, and Perplexity / Sonar providers for wider coverage and answer-first routing.
Concise communication mode. Cuts ~60-70% of output tokens while keeping natural, readable English. Drops filler, hedging, and pleasantries but maintains grammar and sentence flow. Elaborate on request -- ask for detail and get it, then auto-return to concise. Use when user says "concise mode", "be concise", "less verbose", or invokes /concise. Also triggers when user requests token efficiency with readability.
UAT THẬT cho một bản framework sắp phát hành - dựng một dự án TRỐNG hoàn toàn mới, cài overstack đúng đường NGƯỜI MỚI (curl bootstrap từ GitHub raw, không phải file:// working-tree), rồi kiểm chứng năng lực MỚI của chính bản đó có tới tay người dùng không (skill reachable, rule CẮN thật, docs khớp). HAI PHA - (1) nhánh CANARY trước merge: đẩy lên nhánh tạm uat/<ts>, curl từ raw của chính nhánh đó (kèm HARNESS_BASE + REPO_RAW + SKILLS_REF trỏ canary), UAT đầy đủ; FAIL thì xoá canary và nhánh chính chưa hề bị bẩn. (2) main-URL smoke NGAY SAU merge: chạy đúng lệnh người mới gõ, KHÔNG override biến nào - chỉ pha này kiểm được các giá trị MẶC ĐỊNH (chỗ hardcode tên nhánh); FAIL thì gỡ commit khỏi remote (revert hoặc reset --force-with-lease). Gọi khi user nói "uat", "test thật", "test downstream", "kiểm tra người mới cài có chạy không", "acceptance", "/fdk-uat", hoặc trước khi công bố một bản có năng lực mới. KHÁC medic --ci và fresh-install-smoke (cả hai cài từ working-tree qua file:// nên KHÔNG chứng minh được đường remote).
Keep a written evidence ledger and delegate whole-repository searches during investigations spanning many files. Use for debugging, audits, and archaeology when the relevant code location is unknown. Do not use when the relevant files are already identified or for ordinary edits to known code.
Guide to creating and building high-quality MCP servers that let LLMs interact with external services through well-designed tools. Use when the user wants to build, develop, or implement an MCP server integrating external APIs or services in Python (FastMCP) or Node/TypeScript (MCP SDK). To only install/configure an existing MCP server (e.g. Jira), use setup instead.
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.
Mở rộng một draft SPEC ĐÃ ĐƯỢC DUYỆT thành kế hoạch thi hành được — file `DDMMYY-<tên>-PLAN.md` mà một agent KHÔNG có context nào (CLI rẻ chạy headless, không hỏi lại được) vẫn làm đúng: đường dẫn file chính xác, khối Interfaces (Consumes/Produces) khai chữ ký cho task hàng xóm, ràng buộc bao trùm chép nguyên văn, và từng bước 2-5 phút kiểu TDD có code thật + lệnh chạy + output mong đợi. Gọi SAU `/propose` và SAU khi user duyệt ở cổng, TRƯỚC khi dispatch task cho agent. Trigger - "viết plan", "mở rộng proposal thành plan", "plan thi hành", "chuẩn bị brief cho agent", "/plan". KHÔNG dùng để thiết kế hay để hỏi yêu cầu (đó là `/propose`), KHÔNG dùng khi chưa có SPEC duyệt.
Test your AI agent with simulation-based scenarios. Covers writing scenario test code (Scenario SDK), creating platform scenarios via the `langwatch` CLI against a connected agent, reading the run parameters that agent declares so the scenarios and comparison runs turn its real levers, and red teaming for security vulnerabilities. Auto-detects whether to use code or platform approach based on context.
This skill should be used when the harness, scaffold, workflow, or optimizer itself is the optimization target: recursive self-improvement (RSI) loops, meta-harnesses, self-improving harnesses that mine their own failures and propose bounded edits, evolutionary or population-based search over agent scaffolds, acceptance gates for self-modifying systems, and agentic context evolution where the mechanism that produces context is versioned and evolved. Route governance of a single autonomous loop (locked surfaces, durable logs, rollback, novelty gates, approval boundaries) to harness-engineering, measurement and quality-gate design to evaluation, judge design to advanced-evaluation, and remote sandbox infrastructure to hosted-agents.
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.