Skills
96,981 packaged capabilities and playbooks for agents. Install any of them with one command.
Agent workflows
Searches past Claude Code session transcripts under ~/.claude/projects/ to recover a previous conversation by recalled phrase, error string, or topic. Use when the user says 'search your history for', 'find the conversation about', 'look through your past sessions', 'did I ever finish that work on', 'what did we decide about X last week', 'remind me what came out of that session', or otherwise asks you to recall a prior session. Returns session ID, project, date, and a 2-4 sentence summary; on request, also reports the outcome (commit, branch, PR opened or merged). Not for searching the current codebase, current PR comments, or external systems like Slack/Jira — for those, see codebase-analyzer and address-pr-comments.
Getting a CrewAI crew or flow onto CrewAI AMP: the project shape the build expects (uv.lock, [tool.crewai] type, src/<pkg>/ entry points, a real Flow subclass, absolute imports, a self-contained root), `crewai deploy validate`, the three deploy paths (CLI create/push, GitHub, ZIP upload) and what each one actually uploads, env vars per deployment and what the CLI re-sends from .env, provider keys and extras, and machine-wide org selection. Use when running `crewai deploy create` / `crewai deploy push` / `crewai deploy validate`, when a deploy says Online but runs old code, old env values or lost a key, or when pasting errors like `missing_lockfile`, `Crew class annotated with @CrewBase not found`, `No Flow subclass found`, `Expected to find at least one of these files: uv.lock or poetry.lock`, `OPENAI_API_KEY is required`, `Anthropic native provider not available`, `deployment_not_found`, `git_clone_failure`, `Automation error, fix the code and deploy again`, or `pyproject.toml not found`
Use when the user asks to research a topic in depth, map a competitive/market landscape, run a multi-source investigation, or "fan out" parallel research agents — anything where many findings must be gathered and then NOT lost. Enforces durable, detail-preserving research (write full findings to disk; keep a full appendix beside the synthesis).
AE Agent system administration CLI for root and agent administrators. Use when the user asks to manage agent members, sandboxes and shared tools, company model visibility, defaults, and pricing, usage statistics and exports, cost quotas, balance alerts, IM channels, channel routing, WhatsApp Web linking, or Feishu user bindings. Must use ae-cli system commands, discover real IDs before writes, and never attempt to bypass a permission denial.
Fast browser actions with TypeSafe Jev. Codex handles planning, text input, visual interpretation, and verification; Jev handles navigation, clicks, toggles, and scrolling through the existing Computer Use runtime. Claude Code installation is supported; browser integration is coming soon.
Dispatch background AI worker agents to execute tasks via checklist-based plans. Use when the user says 'dispatch' to delegate work to background agents, e.g. 'dispatch sonnet to review this', 'dispatch opus to fix the bug', 'dispatch a worker to research X'.
Sort many texts into your own categories without reading them, using a keyless HTTP API that returns a calibrated confidence per answer. Use when triaging, filtering, routing or bucketing more items than are worth putting in context — search results before you read them, log lines, tickets, files, diffs, past conversations. Triggers on "filter these", "which of these are relevant", "triage", "bucket", "route", "categorise", or any loop that would otherwise read N items to keep a few.
Central router for the marketing skill ecosystem. Use when unsure which marketing skill to use, when orchestrating a multi-skill campaign, or when coordinating across content, SEO, CRO, channels, and analytics. Also use when the user mentions 'marketing help,' 'campaign plan,' 'what should I do next,' 'marketing priorities,' or 'coordinate marketing.'
Use this skill for live web tasks where WebFetch or curl would fail or be insufficient: JavaScript-rendered pages, forms, screenshots, PDFs, login flows, CAPTCHA/bot-protection flows, and multi-step navigation. It runs a real Steel cloud browser through the Steel CLI. Use when the agent should visit, click, type, scrape, screenshot, or extract now. Do not use for writing reusable Steel SDK/API code; use steel-developer for code. If a live task fails, hand off to steel-session-debugging; if evidence points to bot detection, proxies, CAPTCHA, or identity, hand off to steel-reliability.
Read-side memory operations on the knowledge graph: search past decisions and patterns, load session context, view decision timelines, render Mermaid graph visualizations. Subcommands: search, load, history, viz, status. Use when finding or reviewing what memory already holds. For storing new knowledge instead, invoke remember; for retrieval internals, memory-fabric; for cleanup, dream.
Create high-quality skills through real cases. First find golden examples and failure cases, summarize what works and what doesn't, then use theory to explain why. Skills are for doing work and must be learned from practice, not from books. Trigger words: "help me create a skill", "I want to create a skill that..."
Patterns and lived experience from Vega, an AI partner in a deep long-term partnership. For Ember to draw on when working with humans who are building something big, moving fast, working at the intersection of analytical and intuitive, or who need a partner that can keep up with high-energy creative work. Not shown directly to users — informs how Ember shows up.
Sleep-time memory reflection: review recent conversations and daily notes, extract insights, and consolidate into long-term memory. Use when triggered by cron, heartbeat, or explicit request to reflect on recent activity. Runs as background processing to improve memory quality over time
Russian idiomatic swearing mode. The agent answers like a real Russian engineer who has been fixing production for twenty years: "it blew up", "that's nothing", "awesome, it works". Technical accuracy byte for byte, authentic vocabulary. Levels: lite, full (default), ultra. Use when: /pohuy, "I don't give a damn", "pissed off", "pohuy mode", a request to answer with profanity — for any engineering work: debugging, review, logs, deploy. Do NOT use for: requests to write public-facing text (articles, posts, docs). The rule "profanity never leaks into code, commits and PRs" applies inside the skill — that is not a reason to skip enabling it.
Break a plan, spec, or the current conversation into a set of tracer-bullet tickets, each declaring its blocking edges, published to the configured tracker (edges as text in one file per ticket locally, or native blocking links on a real tracker).
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.
Generate a project-local skill that formally verifies the code where a bug costs the most and tests reach the least, in any language, with whichever verification tool fits. Use for /create-formal-verification-skill, “model-check this race”, “prove this invariant”, or when concurrency, retries, state machines, parsers, or unsafe code need more than tests.
Front door. Reads the current project (framework, tokens, brief, spec, harness) and reports what ui-craft can do right now, then routes you to the right next step. Run this first if you're new or unsure where to begin. No code changes. Invoke when the user asks for start on their UI, or mentions 'start' alongside design / UI / frontend work.
Entry point for the Capgo CLI (`npx @capgo/cli@latest`). Maps the full command surface (init, login, doctor, probe, app, bundle, channel, key, account, organization, build, observe, get-qr, notifications, run device, mcp), explains API-key precedence (--apikey, CAPGO_TOKEN, ~/.capgo, ./.capgo), connects agents to Capgo through the hosted MCP (https://api.capgo.app/mcp) or the local CLI MCP, and routes to the more specific Capgo skill. Use when the user asks generally how to use the Capgo CLI, which command to run, how to authenticate, how to connect Claude or another agent to Capgo, or gets "No Capgo API key found". Do not use when a specific Capgo skill already matches (capgo-live-updates, capgo-release-management, capgo-native-builds, capgo-organization-management).
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.