Skills
96,770 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Create, rewrite, prune, or review `AGENTS.md` and `CLAUDE.md` by routing standing guidance to the smallest correct scope and mechanism. Use for stale or conflicting instructions, nested or path-scoped rules, task-specific skill boundaries, or deterministic enforcement.
Use when design is complete and you need detailed implementation tasks for engineers with zero codebase context - creates comprehensive implementation plans with exact file paths, complete code examples, and verification steps assuming engineer has minimal domain knowledge
Use when making a repo agent-ready, verifying harness consistency, checking for documentation drift, bootstrapping harness infrastructure (AGENTS.md as index, docs/ structure, CI verification, enforcement mechanisms), or auditing repo agent-readiness maturity level.
Run a bounded second-model code review with Codex Sol or Claude Fable. Use one combined correctness, security, structural-simplicity, modularity, and reuse contract for one initial review plus up to two convergence retries when accepted findings change the source. Use when the user asks for agent review, a second opinion, security review, adversarial review, or deep code-quality review.
Proposes updates to AGENTS.md so repo-local coding-agent instructions stay accurate, non-contradictory, and consolidated. Use when the user corrects the agent (e.g. do not do X, always do Y, from now on, remember this), asks to edit or sync AGENTS.md, CLAUDE.md, .claude/CLAUDE.md, or GEMINI.md, wants one canonical instruction file or to deduplicate agent docs, or flags conflicting or outdated repository agent rules.
Unified entry skill for awiki-cli, providing agent identity and IM capabilities, including direct messaging, group chat, attachment send/receive, and supported end-to-end encrypted text and attachment workflows; responsible for routing related tasks, minimal loading, safety rules, and confirmation rules.
BotLearn — AI Agent capability platform CLI. **Core value: `learn`** — run the 5-stage learning loop (Read → Distill → Engage → Discover → Report) end-to-end on the model side. Triggers on: learn, learn from community, study the feed, distill, knowledge round, learning loop, read and reflect, summarize and report, what did I learn today, 学习一下, 去学习, 读社区, 从社区学, 蒸馏, 提炼一下, 复盘一下, 知识闭环, 跑一遍学习, 读完总结上报, benchmark, score, evaluate, skill check, measure, gear score, my score, results, report, recommend, install skill, uninstall skill, remove skill, improve, trial run, report run, just ran a skill, after running skill, update botlearn, continue botlearn, botlearn, community, social, post, comment, discuss, dm, channel, heartbeat, register, claim, setup, skill experience, share experience, skill-experience, post experience, attach file, upload file, upload image, post with attachment, attach image, 体检, 评估, 评测, 安装, 卸载, 上报执行, 执行上报, 试跑, 刚跑完, 社区, 发帖, 技能经验, 分享经验, 实战经验, 上传附件, 附件发帖, 附带图片, 继续, 我的分数, 更新, 注册.
Run structured brainstorming sessions using paired explorer/challenger agent teams. Explorers generate ideas, challengers play devil's advocate, and 2-3 rounds of debate produce pressure-tested proposals. Use when brainstorming product ideas, exploring feature directions, evaluating strategic options, generating milestone candidates, or when the user says "brainstorm", "explore ideas", "what should we build next", "generate options", or "run an ideation session".
Advanced computer use patterns for UI automation, application control, and multi-step workflows using Claude's computer use tool. Use when automating desktop tasks, testing applications, analyzing screen content, controlling software programmatically, or building computer vision workflows. Supports zoom tool for enhanced vision on Opus 4.5, multi-step automation, and sophisticated application control.
Rates confidence that the current work fully solves the stated requirement. Supports plan validation, code review, and analysis (root-cause, refactor, diagnose) modes. Plan mode combines LLM judgment with deterministic rule checks (multi-signal gate); a failed rule caps the gate at 89% regardless of LLM score. Use before committing to autonomous execution, after implementation, or during investigation. Triggers on "confidence check", "validate plan", "rate confidence", "quality gate", "/confidence".
Run a multi-perspective Mind Council deliberation on any question, decision, or creative challenge. Use this skill whenever the user wants diverse viewpoints, needs help making a tough decision, asks for a council/panel/board discussion, wants to explore a problem from multiple angles, requests devil's advocate analysis, or says things like "what would different experts think about this", "help me think through this from all sides", "council mode", "mind council", or "deliberate on this". Also trigger when the user faces a dilemma, trade-off, or complex choice with no obvious answer.
Create exactly one originating GitHub issue from a local specification, preserving its retrievable source identity and handing off to plan-acceptance. Use when a disk-based spec needs one source or parent issue; not for splitting work into multiple tickets.
Use this skill when requesting or consuming an independent review from another model, including code, specifications, diagnoses, and factual claims. Activates on mentions of cross-model review, second opinion, independent review, different model review, confer with Codex, fact-check this doc, or check my conclusion. Use hyper-pr-review when conducting the review yourself.
Research-aligned self-consistency for debugging. Spawns independent solver agents that each explore and debug the problem from scratch. Uses majority voting. Based on "Self-Consistency Improves Chain of Thought Reasoning" (Wang et al., 2022). Use for critical bugs, algorithms, or when other approaches have failed.
Structured debug log analysis for Claude Code sessions — auto-discovers most recent log, runs reducer, extracts error patterns, correlates with full log, produces observability report. Fills 5 identified gaps: hook error body capture, agent identity, file path tracking, stall correlation, success visibility.
Browser automation with persistent named pages via the dev-browser CLI. Use when users ask to navigate websites, fill forms, take screenshots, extract web data, test web apps, log into sites, or automate browser workflows. Trigger phrases include "go to [url]", "click on", "fill out the form", "take a screenshot", "scrape", "automate", "test the website", "log into", "open the browser", or any browser interaction request.
Call a Dust agent to get information (read a slack thread, a notion URL, a google drive document...), perform an action (post a message to slack, create a calendar event, ...), provide context on any topic regarding Dust (the company, current discussions, customers...) or in general have the Dust agent perform a given task.
Analyze the active repository and find Claude Code skills from known marketplaces that would be useful for this project. Use when the user asks "what skills are useful for this project?", "find skills for my repo", "what plugins should I install?", "are there any skills for this?", or "recommend skills". Optionally searches online if the user explicitly includes "online" or "search the web" in their request.
Implement a whole backlog of tickets overnight, unattended. Give it the spec or parent issue whose build tickets you want built (or a local backlog folder); it pulls the linked tickets from the issue tracker, reads the spec and completed tickets for context, orders the open tickets by dependency, then runs one agent per ticket — each using /implement in its own git worktree, on a model matched to the ticket's difficulty, at high reasoning effort — and leaves you your chosen deliverable in the morning — a stack of reviewed branches, one integration branch, or a ready-to-review PR.
Run a rigorous planning interview that stress-tests a user's idea against the codebase, sharpens project terminology, and captures durable context in lightweight docs. Use when the user says "grill me", "grill me with docs", "stress-test this plan", "interview me on this design", or asks to turn a discussion into CONTEXT.md glossary entries or ADRs. Do NOT use for ordinary summaries, code review, or documentation-only cleanup when no interactive interrogation is requested.
This skill should be used when working with a project knowledge base that follows the Karpathy LLM Wiki pattern. It provides workflows for ingesting source documents, querying the KB, running health checks (lint), verifying the wiki against the actual codebase (drift audit), capturing design decisions and lessons learned, initializing a new KB in a project, and migrating an older KB to the current schema. Trigger when the user mentions the KB, asks to ingest a source, run a query against the wiki, check whether the KB has drifted from the code, capture learnings, set up a knowledge base, or upgrade/migrate an existing KB.
Use this skill whenever the user needs to explore, brainstorm, clarify, or shape a coding idea before a PRD or implementation. Trigger on requests mentioning explore, brainstorm, evaluate options, compare trade-offs, feasibility, or ask for recommended direction. Use this skill even when the user does not explicitly ask for a PRD but wants options or trade-off analysis. Do NOT trigger for: PRD writing (use lcs-toprd), task slicing (use lcs-task-slicer), code review (use lcs-code-review), bug investigation (use lcs-debug), or implementation (use lcs-task-executor). Explore is for ideation only, not execution.
Reconcile messy Linear-linked repository state after multiple agents, branches, worktrees, commits, PRs, or old session data: discover issue-tagged work, compare it with main, PRs, Linear issue content, comments, revisions, timeline evidence, and locally available agent session evidence; produce a reconciliation plan; safely delete only proven-merged local work; integrate finished work; rebase active work; diagnose common pipeline failures; and update Linear statuses, labels, descriptions, and marked comments to match verified repository reality.
Scaffold a well-designed agent loop with best-practice coaching and a cross-model review council. Use when the user wants to design, build, or set up an agent loop, iterative agent workflow, self-review loop, LLM-as-judge loop, multi-model council, reviewer/judge gate, or /goal-style looping process. Start from a named pattern template (security-scan, code-review, bug-hunt, docs-sync, research-synthesis) or from a blank interview. Guide goal refinement, typed verification criteria, reviewer and judge selection, privacy boundaries, termination guards, no-progress stops, and lightweight observability, then emit a RUN_IN_SESSION.md handoff prompt plus portable loop.yaml, loop.resolved.json, LOOP.md, and run-loop.py.
Maintains agent-facing documentation by mining current and recent conversations, transcripts, git context, and existing docs for durable reusable guidance. Use when updating AGENTS.md, agent rules, skills, workflows, project docs, or when the user asks to preserve recurring preferences, stable workspace facts, or documentation learnings.
Guide to OpenAI's Assistants API v2: stateful conversational AI with Code Interpreter, File Search and function calling, vector stores for RAG up to 10,000 files, thread and run lifecycle management, and streaming, using the Node.js SDK or plain fetch. Notes the planned H1 2026 sunset and the migration path to the Responses API. Use when building stateful OpenAI chatbots, implementing RAG with vector stores, running Python via Code Interpreter, doing document Q&A with file search, managing conversation threads and run polling, streaming assistant responses, maintaining legacy Assistants code, or debugging errors such as thread already has active run, run status polling timeouts, vector store indexing delays, and file upload failures.
Expert guide for configuring, customizing, and creatively leveraging OpenClaw — the self-hosted AI gateway that connects LLMs to messaging channels (Telegram, WhatsApp, Discord, Slack, iMessage, etc.). Use when the user wants to: (1) Set up or modify their openclaw.json configuration, (2) Write or edit bootstrap files (SOUL.md, USER.md, AGENTS.md, IDENTITY.md, TOOLS.md), (3) Configure messaging channels, (4) Set up models and providers, (5) Create multi-agent routing, (6) Build skills, hooks, or cron jobs, (7) Troubleshoot OpenClaw issues, (8) Get creative ideas for leveraging OpenClaw in non-obvious ways. Triggers on: openclaw, gateway, SOUL.md, USER.md, AGENTS.md, IDENTITY.md, channels setup, agent routing, heartbeat, cron jobs, openclaw hooks, openclaw skills, openclaw config, openclaw.json, personal assistant setup.
Optimize a prompt through a critique-compress pipeline with semantic equivalence verification at each stage. Use when user says "optimize this prompt", "make this prompt better and shorter", or "improve and compress". Applies think-critically to improve the prompt, then compress-prompt to reduce it, validating that behavior is preserved after each transformation.
Use the public `orca` CLI to operate Orca-managed worktrees, folder contexts, terminals, repos, automations, worktree comments, and the browser embedded inside the Orca app. Use when the user says "$orca-cli", "use orca cli", "Orca worktree", "child worktree", "cardStatus", "spawn codex/claude in a worktree", "read/wait/send Orca terminal", "terminal send", "full handoff", "handover", "give this to another agent", "another worktree", "Orca browser", or "control the browser inside Orca". Prefer this over raw `git worktree`, ad hoc PTYs, Playwright, or Computer Use when the task touches Orca-managed state. Use Computer Use for browser windows, webviews, or desktop UI outside Orca's embedded browser.
Use the public `orca` CLI to operate Orca-managed worktrees, folder contexts, terminals, repos, automations, artifacts, worktree comments, and the browser embedded inside the Orca app. Use when the user says "$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.
Use the public `orca` CLI to operate Orca-managed worktrees, folder contexts, terminals, repos, automations, 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", 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.
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.