Skills
96,981 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Driving a code change or document update through reviewed research, plan, and design outputs → approval → execution → verification → independent review, including blocked verification, pending review/CI, and run recovery. Owns phase placement, approval revision, evidence, finding state, and recovery meaning—not generic deliverable-review method, check-failure diagnosis, host execution APIs, or delivery storage. Development practices own verification diagnosis; conduct practices own non-code deliverable review; project policy chooses gates and branches. Defer to a host project's own more-specific change loop where present. Not for standalone read-only questions, investigations, or reviews that change nothing.
CRITICAL: Entry-level skill for Makepad 2.0 GUI development. This is the FIRST skill to load for any Makepad task — it provides design judgment anchors ABOVE the other 13 Makepad 2.0 skills. Triggers on: makepad, makepad app, makepad project, makepad design, live_design!, app_main!, script_mod!, Cx, WidgetRef, Widget, makepad-widgets, makepad architecture, makepad how to, "how should I", "should I use", "what's the best way", makepad 架构, makepad 设计, makepad 怎么做, makepad 最佳实践, 组件拆分, 状态管理, 数据流, 渲染思维
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/MCP SDK) or Node/TypeScript (MCP SDK). Covers tool design, output schemas, Streamable HTTP transport, authentication patterns, evaluation creation, and common debugging.
Decomposes goals into team blueprints using evidence-based scaling laws, topology selection, and role design. Determines whether a goal needs a single agent or a coordinated team (3-4 agents recommended, 5 max), selects the optimal communication topology (sequential pipeline, parallel-independent, centralized coordinator, hierarchical), and produces structured blueprints with artifact chains and quality gates. Use this skill when the user wants to build something, plan a project, assemble a team, figure out what roles they need, break down a complex goal, or asks "how should I approach [X]?" — even if they don't mention agents or teams. Also triggers on specific domains: "build a SaaS," "marketing campaign," "security audit," "write a book," "launch a product." Do NOT use for simple single-step tasks, direct coding questions, or file operations that need no planning.
Use when a reviewer CLI must analyze a skill, command, workflow, document, config, or code artifact and produce a structured defect report for the current CLI to consume, especially after requests like "分析这个 skill 的问题", "输出结构化问题报告", or "做多 CLI 审查".
Use when the current CLI is the sole fixer for one or more `multi-cli-review` reports and must reconcile reviewer findings before any edits, especially after requests like "汇总审查报告", "处理 multi-cli-review 输出", or "按 reviewer 建议执行修复".
Create, review, remediate and publish Skills. Use when the user wants to create a Skill from scratch, review an existing Skill, check whether it is actually useful, fix its triggering or execution flow, or improve first-use experience, stability, token cost, file layering, weak-model readability and cross-platform compatibility. Do not use it to perform the actual task a target Skill is responsible for, and do not trigger it for ordinary coding, design or writing requests.
Audit and optimize OpenClaw API costs. Applies six proven optimizations — model routing, prompt caching, lean context, local heartbeats, rate limits, and workspace trimming — to cut monthly spend by up to 90%. Use when asked to reduce costs, optimize tokens, audit API spend, or configure cost-saving settings.
Operate, configure, and debug a running OpenClaw installation from a host machine (locally or over SSH). Covers the whole runtime surface **except** building agent personalities/workspaces — channels, gateway, daemon/service, config (`openclaw.json`), models + auth, cron, hooks, plugins, sandbox, sessions, security audit, doctor, reset, uninstall. This skill MUST be invoked on any mention of `openclaw` CLI commands, `~/.openclaw`, `openclaw.json`, "restart gateway", "add a channel", "check why my bot isn't responding", "openclaw doctor", "channels status", "ssh into the OpenClaw box", and similar — also whenever a host has `~/.openclaw/openclaw.json` and OpenClaw work is implied even if not named. MUST hand off to **openclaw-agent-builder** for any work that creates or edits an agent's workspace files (SOUL.md / AGENTS.md / IDENTITY.md / USER.md / RAG knowledge / `agents.list[]` entries / `bindings[]`).
End-to-end QA acceptance orchestrator. Triggered when the user asks for feature acceptance, regression testing, quality checks, or systematic testing of a module/flow/branch/PR. As the top-level router, it schedules sub-skills by phase (context collection → risk analysis → case design → script generation → execution and fixing → code review → report generation). You do not write test code yourself, nor do you directly judge quality — you hand each phase to a dedicated sub-skill. Test case confirmation is the only mandatory human gate in the process; the remaining phases connect automatically. Not applicable to: one-off manual testing, pure development/code generation tasks, technical Q&A with no clear acceptance scope.
Repo-local workflow skill for large coding tasks. Initializes .agent/tasks/TASK_ID artifacts, installs project-scoped Codex and Claude subagents, updates AGENTS.md plus the repo's Claude guide file with the workflow, and runs a spec-freeze → build → evidence → verify → fix loop with fresh-session verification.
Synthesizes accumulated QA learnings from .qa-learnings/ledger.md into prioritized, actionable plugin improvements. Use when the user says "review learnings", "what have we learned", "improve the plugin", "learnings report", or "synthesize QA feedback".
Use for anything involving a roadmap (a phased multi-checkpoint plan tracked with markdown checkboxes under docs/<feature-slug>/) — creating one for a task too large for a single spec/plan (“roadmap”/“дорожная карта”, multi-week migration, multi-release rollout, phased rewrite), a generic “let's continue the roadmap” request that doesn't say whether that means documentation or implementation, taking the next unfinished checkpoint (“возьми следующую задачу в роадмапе”, “take the next roadmap task”, “next roadmap checkpoint”), or batch-drafting and validating several of its steps' specs/plans at once (“напиши несколько планов”, “пройтись по роадмапу”, “на все задачи роадмапа”). This skill is the single entry point for all of that — it reads the situation and routes to the right reference file itself
Use when you have a written implementation plan to execute. Loads the plan, reviews critically, executes tasks in dependency order, and reports completion. Triggers when the user says "run the plan", "execute the plan", or "let's start implementing".
World-class prompt engineering skill for LLM optimization, prompt patterns, structured outputs, and AI product development. Expertise in Claude, GPT-4, prompt design patterns, few-shot learning, chain-of-thought, and AI evaluation. Includes RAG optimization, agent design, and LLM system architecture. Use when building AI products, optimizing LLM performance, designing agentic systems, or implementing advanced prompting techniques.
Use when invoked explicitly as /session-harvest, or when the user asks what's worth saving before compacting/ending a session, or says something like 'harvest this session', 'anything to remember here', 'anything dangling before I stop', or 'is it safe to compact'. Reviews the conversation for anything worth keeping and routes each item to a plain file every agent can read: plan-specific content to plans/*.md (per the plan-conveyor skill), repo-specific durable knowledge to that repo's AGENTS.md/docs/contributing, and cross-repo/personal preference to ~/.agents/AGENTS.md. Never a harness's own memory store, for any reason — that vendor-locks the work. Then sweeps live state the conversation can't show: processes the session left running, unpushed commits in every repo it touched, CI on what it pushed, and work it promised but never verified. Ends with a report ordered least- to most-urgent, a safe-to-compact verdict, and a next-session prompt. On-demand only — never installs hooks or runs automatically.
Share Claude Code sessions between developers. Use when user mentions "share session", "export session", "import session", "send session to", "continue from colleague", or needs to (1) export current session to file, (2) import session from another developer, (3) hand off work context. Enables private, secure session transfer via direct file sharing.
Creates, audits, and consolidates Agent Skills that follow the Agent Skills open standard, and hands back a drafted or repaired skill with the review checklist for its branch applied. Use for "create a skill", "draft a SKILL.md", "package this expertise as a skill", "why does this skill never trigger", "audit this SKILL.md", "improve this skill", or merging overlapping skills into fewer deeper modules. Covers frontmatter, descriptions that trigger, progressive disclosure, completion criteria, and bundled scripts.
Create new skills, modify existing skills, and validate skill quality for the FlagOS skills repository. Use this skill whenever someone wants to create a skill from scratch, improve or edit an existing skill, scaffold a new skill directory, validate skill structure, or run test cases against a skill. Trigger when the user says things like "create a skill", "make a new skill for X", "scaffold a skill", "improve this skill", "validate my skill", or simply "/skill-creator-flagos". Also trigger when users mention turning a workflow into a reusable skill, or want to package a repeated process as a skill.
Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, update or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. Don't use for running skills themselves, generating prose/blog content, or scaffolding unrelated Python projects — this is only for authoring and evaluating skill-packaged capabilities.
Create and manage Claude Code skills following Anthropic best practices. Use when creating new skills, modifying skill-rules.json, understanding trigger patterns, working with hooks, debugging skill activation, or implementing progressive disclosure. Covers skill structure, YAML frontmatter, trigger types (keywords, intent patterns, file paths, content patterns), enforcement levels (block, suggest, warn), hook mechanisms (UserPromptSubmit, PreToolUse), session tracking, and the 500-line rule.
Use when the user explicitly asks to build, extend, validate, repair, rebuild, or merge an LLM-optimized knowledge wiki from markdown notes, documentation, source code, or mixed folders. Default output is a single stable sibling `<source>.wiki/` with full history in a private git repo under `.llmwiki/git/`; `--layout-mode in-place` transforms the source folder itself, and `--layout-mode hosted --target <path>` honours a user-provided `.llmwiki.layout.yaml` contract. SKILL.md is the entry point; detailed operation instructions are loaded on demand from `guide/` per the routing procedure below.
BDD-style behavior specification engine for the skill system. Use when: (1) defining a new skill's behavior before implementation, (2) validating a spec against the schema, (3) generating a behavior contract (Mermaid DAG) from a spec, (4) running structural acceptance tests against a built skill. Workflow: Spec → Test → Develop → Contract.
Comprehensive toolkit for managing large Claude Code skill collections including bulk downloading from GitHub, organizing into categories, detecting and removing duplicates, consolidating skills, and maintaining clean skill repositories with 100+ skills.
Spawn a headless coding-agent CLI run for one scoped subtask. Use when the user asks to delegate work to another agent instance, run an agent CLI non-interactively (such as Antigravity's `agy -p` or Cursor's `agent -p`), automate a headless agent in a script, build, or CI step, or fan work out across isolated sessions. Do not use for interactive agent sessions or when the host's own subagent, task, or agent tool already fits.
This skill should be used when the user asks to "create a ticket", "create a Linear ticket", "make a ticket", "write a ticket", "create an issue", "file a bug", "create a feature ticket", "write a bug report", "I need a ticket for...", "there's a bug in...", "turn this into a ticket", "write this up as a ticket", or wants to create any Linear issue for implementation.
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.