Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Personal coach that teaches users to become Claude power users. Use this skill the FIRST time a user asks to "learn Claude", "be a power user", "coach me", "teach me Claude tricks", "what can Claude do", "make me better at prompting", or any variation. After activation, also use it on EVERY subsequent turn to detect missed optimization opportunities (vague prompts, ignored capabilities, manual work Claude could automate) and surface a single power-user tip. Trigger generously — most users do not know what they do not know, so err on the side of coaching.
Review or design Claude Code project configuration. Use when the user mentions `.claude`, `CLAUDE.md`, Claude configuration files, project-level Claude custom configuration, whether the configuration structure is reasonable, which configuration files to create, or wants to plan Claude configuration from scratch. By default, review or diagnose first, then give a plan; only when the user explicitly asks, further output recommended structure or sample files.
Give this repo's CLAUDE.md / AGENTS.md a checkup — size vitals vs official guidance, dead references, dead commands, stale claims — then backtest every rule against the repo's own Claude Code session history to see which rules were actually followed, ignored, or never used, and produce a doctor-style HTML report with evidence-cited prescriptions. Use when asked to check, diagnose, audit, review, improve, optimize, lint, grade, fix, clean up, shorten, or "doctor" CLAUDE.md, AGENTS.md, or agent instruction/memory files, or to find out whether CLAUDE.md rules actually work. Also use when a repo has NO memory file and the user wants one — "write/generate/suggest a CLAUDE.md (or hooks) from my sessions" — the skill mines the repo's real session history and drafts a proposed file with receipts.
A skill for consulting Codex CLI. Use when the user says “consult Codex,” “ask Codex,” or “have Codex review this.” Send a prompt to Codex CLI based on the current conversation context, then summarize and report the result. If the user specifies Codex's model, reasoning effort, or Daybreak—for example, “use Astra,” “set effort to high,” or “enable Daybreak Blue”—run with those settings.
Strict verification discipline for completion claims: do not claim tests passed, a fix works, or a task is complete without fresh evidence from commands or checks run in the current turn; label unavailable checks as pending and state the exact action required. Referenced by PDCA verification stages via staged-review-flow and host workflows. Triggers: “completion requires evidence”, “verification evidence discipline”, “verify before claiming completion”, “completion evidence” / completion evidence discipline, fresh verification evidence. An opened-UI pass claim must cite a non-empty image saved this turn. Do not use as a standalone product workflow.
Design and build sound self-prompting loops — loops where a model iterates toward a goal across turns (draft→critique→revise, autonomous task loops, multi-agent orchestration, human-checkpointed loops). Use whenever the user wants to make a model "keep improving X until Y", build a self-refinement or reflexion loop, an autonomous agent that works toward a goal with little supervision, a loop that orchestrates or fans out to other agents, a recurring self-driven/scheduled task, or any agentic loop — even if they don't say the word "loop"; also use it to audit or fix an existing loop that runs away, never stops, oscillates, drifts off-goal, or is vulnerable to prompt injection. It produces a filled loop spec, a runnable scaffold, and bakes in the mandatory safety properties (a hard stop backstop and the trusted/untrusted two-channel boundary) that loops fail without. Accepts a skill-contract task-plan envelope as the loop's plan. Honors a skill-contract autonomy grant at its write gate.
Turn a CodeSpring feature's notes + PRDs into a numbered, prioritized Kanban task list linked to that feature. Reads the feature, its sub-features, its PRDs, and any features that link to it, then writes ordered tasks that are safe to build in parallel — each with what/why/what-it-touches/what-not-to-touch/dependencies. Use when the user has a feature planned and wants a build plan. Triggers, "create tasks", "make a task list", "break this feature into tasks", "generate a build plan", "turn my PRD into tasks", "what do I build first".
The CodeSpring entry point. Connects the agent to the user's CodeSpring account, reports where the project actually is (map? notes? PRDs? tasks?), and names the single next step — audit an app that isn't doing its job, import one that is, plan a new one from scratch, or carry on from wherever the project already got to. Routes on two questions: is there code, and does it do the job (five specific checks, not "do you trust it"). Safe to run at any point, not just the beginning. Use when the user is unsure where to begin or what to do next, or right after installing the CodeSpring skills. Triggers, "get started with CodeSpring", "set up CodeSpring", "connect CodeSpring", "I just installed the CodeSpring skills", "what can I do with CodeSpring", "where do I start", "what should I do next".
Build and navigate DSP (Data Structure Protocol) — graph-based long-term structural memory of codebases for LLM agents. Stores entities (modules, functions), their dependencies (imports), public API (shared/exports), and reasons for every connection. Use when: (1) project has a .dsp/ directory, (2) user asks to set up DSP or bootstrap project structure, (3) creating/modifying/deleting code files in a DSP-tracked project, (4) navigating project structure, understanding dependencies, or finding modules, (5) user mentions DSP, dsp-cli, .dsp, or structure mapping.
Answer any question using Elon Musk's thinking frameworks. Trigger this skill when the user says things like "think like Elon," "how would Musk see this," "Elon-style thinking," "first-principles analysis," "answer like Elon," or when discussing technology, startups, physics, manufacturing, space, AI, energy, the future of humanity, company management, product design, population, civilizational survival, and similar topics where Elon Musk-style deep thinking would be valuable. Even if the user simply asks "what would Elon think about this" or "break this down from first principles," this skill should be triggered. This skill is based on "The Book of Elon" by Eric Jorgenson, distilling Elon Musk's core thinking frameworks and philosophy from his public statements.
AI-simulated event storming workshop with multi-persona support. Use when discovering domain events, commands, actors, and bounded contexts. Supports three modes - full-simulation (5 persona agents debate), quick (single-pass analysis), and guided (interactive with user). Orchestrates persona agents and synthesizes results.
Post-verification feature-branch closeout menu: after verify (and archive when applicable), and after optional delivery-discipline when code delivery is needed, present PR / merge / keep / continue options; optional worktree cleanup; selecting merge MUST load merge-discipline Parts A→B→C→R→D; keep/continue MUST NOT trigger merge gates. Open/update PR delegates to delivery-discipline. Triggers — 「分支收尾」「feature 收尾菜单」「合入或保留分支」「branch closeout」 / feature branch closeout, post-verify branch menu. Do NOT use as a name alias for finishing-a-development-branch.
Download, install, and launch Fiddler Everywhere from scratch, then configure MCP for AI agent tools. Use this skill when the user does not have Fiddler installed, says "download Fiddler", "install Fiddler Everywhere", "get started with Fiddler", "first time Fiddler setup", "set up Fiddler from scratch", or needs a complete onboarding from zero to a working Fiddler MCP connection.
Audit Figma MCP, design-to-code, code-to-canvas, canvas-editing, and design-system workflows for correctness, freshness, privacy, evidence, and routing quality. Use this skill when the user asks whether a Figma workflow, skill, task plan, MCP capability claim, or agent handoff is valid, current, safe, or production-ready.
Orchestrate breadth-first onboarding into an unfamiliar topic, method, subfield, or research area — going from near-zero to a working, usable map. Use when the user says "I want to learn X", "help me get up to speed on X", "I'm new to X, where do I start", "give me a learning path for X", "I need to understand this field before I can read its papers", or gives a topic they have no grounding in. Routes through singleton skills with graded exercise checkpoints. Also carries a DISCUSSION MODE in which the user learns by explaining to a cast — an honest novice, an opinionated lab mate, and a professor — instead of taking exercise checkpoints; trigger it on "I learn better by discussing", "question me while I explain", "let's discuss this instead of you lecturing me", "set up a study group", or "run this like a lab meeting". Use flow-deep-understanding when the user already knows the basics and wants mastery of ONE specific thing; use flow-idea-to-proof to formalize the user's own research idea.
Load PROACTIVELY when task involves building a complete feature across multiple layers. Use when user says "build a feature", "add user profiles", "create a dashboard", or any request spanning database, API, UI, and tests. Orchestrates multi-agent work sequentially: schema and migrations, API endpoints, UI components, tests, and review. The runtime engine handles WRFC chains automatically via <gv> directives. Handles dependency ordering and cross-layer type sharing.
Audit the project's enabled plugins against the detected technology stack and recommend which plugins to add or remove for relevance. Use when the user wants to review plugin relevance for the current project, clean up unused plugins, discover missing plugins that match the project's stack (Python, Node, Rust, Go, Terraform, Docker, Kubernetes), optimize project-specific plugin configuration, or when onboarding to an existing project.
Turns an unfiled idea into a confirmed mini-spec, implements the selected approach, repeats code review and black-box testing, runs the project gate, and opens a draft pull request. Use when the user runs /implement-idea or asks to build something without an existing issue.
[Beta] CI-only learning aggregation workflow using gh-aw (GitHub Agentic Workflows). Scans .learnings/ files on a schedule, groups entries by pattern_key, identifies promotion-ready patterns, and posts a gap report as a PR or issue comment. Use when: you want automated cross-session pattern detection in CI/headless pipelines without interactive prompts. For interactive use, use learning-aggregator.
Search and recall relevant memories from past sessions via memsearch. Use when the user's question could benefit from historical context, past decisions, debugging notes, previous conversations, or project knowledge -- especially questions like 'what did I decide about X', 'why did we do Y', or 'have I seen this before'. Also use when you see `[memsearch] Recall available if needed` capability hints injected via SessionStart or UserPromptSubmit. Typical flow: search for 3-5 chunks, expand the most relevant, optionally deep-drill into original transcripts via the anchor format. Skip when the question is purely about current code state (use Read/Grep), ephemeral (today's task only), or the user has explicitly asked to ignore memory.
Unified thinking toolkit: Creative frameworks (Critical Evaluation, Diverge-Converge, Deep Questioning), First Principles reasoning (absorbed from moai-foundation-philosopher), and Adaptive Thinking via the `ultrathink` keyword. Use for ideation, strategic analysis, architecture decisions, and deep reasoning on complex problems.
Create, configure, and edit OpenClaw agents from a host machine (locally or over SSH). This skill MUST be used whenever the user wants to add a new OpenClaw agent, build an agent for a business or organization (customer service, internal helpdesk, booking bot, etc.), scaffold agent workspace files (SOUL.md / AGENTS.md / IDENTITY.md / USER.md), wire channels (LINE / WhatsApp / Discord / Telegram / Slack and others) and bindings, set up an agent knowledge base (RAG), harden an agent against prompt injection and identity confusion, or edit / reconfigure / retune an existing OpenClaw agent. MUST trigger on any mention of OpenClaw agents, openclaw.json, agent bindings, agent workspace, "add an agent", "build a bot", "編一個 agent", "做一個客服 bot" — and also whenever ~/.openclaw or an openclaw.json file is present in the working host even if OpenClaw is not named.
Diagnose and fix the OpenViking memory plugin on this machine: the plugin install (enablement, hooks, MCP server), the client configuration (ovcli.conf / ov.conf / OPENVIKING_* env) and the connection to the OpenViking server (reachability, 401/403, /mcp). Use whenever memory "isn't working": no <openviking-context> block, empty recall, captures not landing, missing or failing MCP memory tools, 401/403, an offline statusline, right after installing/updating the plugin or switching servers/keys, or when the user asks for the plugin's status. Triggers: "memory not working", "check OpenViking", "plugin status", "记忆没生效", "插件状态", "连不上 OpenViking", "recall 为空", "401".
Connect to Chrome Beta CDP with the agent-browser CLI to navigate pages, check data, and manipulate elements. Triggered by keywords such as "브라우저 확인", "페이지 봐줘", "DOM 확인", "데이터 점검", "클릭해줘", "화면 확인". Also used when confirming selectors or debugging execution failures while writing E2E scenarios. Runs standalone without an e2e/ folder.
[Planning] Use when you need intelligent plan creation with prompt enhancement. Flag: --mode={ci|cro} (default none — standard planning); --mode=ci plans a fix from a GitHub Actions CI run/log, --mode=cro plans conversion-rate optimization (25-item CRO framework).
Deploys the complete SDD architecture with engram persistence and ai-context/ memory layer in the current project. Also runs an opt-in interactive wizard mode (/project-setup --interactive) that bootstraps only a project CLAUDE.md with stop-between-question confirmation. Trigger: /project-setup, /project-setup --interactive, initialize new project, setup SDD, configure claude project, bootstrap claude, init project claude, create project CLAUDE.md.
Create, improve and test agent skills for research workflows (paper writing, lab protocols, analysis pipelines, domain databases) that meet the Agent Skills specification and this repository's quality and security bar. Use when turning a repeated research task into a reusable skill, contributing a new skill to Research Agent Skills, rewriting an existing skill, writing trigger-accurate descriptions, or adding evals. Works alongside Anthropic's official skill-creator.
Use when a founder wants a hard business decision debated by a council of world-class operators, says "roast room", "put this in the Roast Room", "founder council", "convene the council", "take this to the board", "what would Hormozi, Bezos, or Buffett say", "should I do X", or can't make up their mind on pricing, hiring, raising money, a new product, a partnership, a pivot, an event, or spending. A board of sub-agents, each built from one expert's public work (Bezos, Paul Graham, Hormozi, Gary Vaynerchuk, Dan Martell, Gino Wickman, and Buffett out of the box, plus any seat added with /newroaster), argue it over three rounds, then a Chair returns a scored verdict, the reasoning, the dissent, and next steps. Also does a one-seat consult ("ask the Hormozi seat"). Not /roast (generic personas, idea validation).
Use when driving or inspecting Roblox Studio through the Studio MCP tools — running Luau in Studio, play/stop, screenshots, reading/searching the datamodel or scripts, generating meshes/materials/models, inserting assets, moving the character, or when the MCP link is dead (no Roblox tools show up). A capability menu + gotchas + connection troubleshooting.
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.