Skills
96,770 packaged capabilities and playbooks for agents. Install any of them with one command.
Productivity
Invoke the `slack-cli` binary to read and act on a Slack workspace from the command line — list channels, read/search conversation history and threads, fetch unread messages, search users, manage user groups, post messages, add reactions, mark channels read, and manage saved items. Use whenever a task needs Slack data or actions, such as "what are the unread messages in #incidents", "search Slack for the deploy thread", "who is @alice", "post a message to #general", "list channels matching X", "react with :rocket: to that message", "what did the team say about Y", or "show replies in this thread". slack-cli is the no-daemon CLI alternative to the slack-mcp-server; it reads a shared on-disk cache so it is cheap to call repeatedly. Output is JSON by default, so results pipe cleanly into `jq`.
MANDATORY recipe for every Caffeine build that lists, reads, creates, shares, or organizes files and folders on the user's own Google Drive. The ONLY supported path is the `googledrive-client` mops package (Drive REST API v3) combined with the `google-oauth` mops package (OAuth 2.0 token exchange + refresh + PKCE). Hand-rolling `ic.http_request` calls to `oauth2.googleapis.com` or `www.googleapis.com/drive/v3` is a FORBIDDEN anti-pattern — it bypasses bearer auth, the `is_replicated = ?false` replication-cost safeguard, and the `google-oauth` library's token handling. Load this skill ONLY when the user, spec, or a prior task refers to **Google Drive specifically** — e.g. "Google Drive", "my Drive", "Drive files/folders", a Drive file/folder ID or share link, "upload to Google Drive", "list my Drive files", or Drive sharing/permissions. Do NOT load it for generic file storage, documents, uploads, or access-control features that are not Google Drive — those are unrelated and this connector must not be attached to them. When it does apply, load it BEFORE writing any code that touches a Google endpoint.
Red-team a PRD, roadmap, or strategy by attacking its load-bearing assumptions before reality does. Steelmans then attacks each claim, ranks failure modes by impact × likelihood × cheapness-to-test, and returns the cheapest test and kill criteria for each. Use when stress-testing a plan, pressure-testing a strategy, challenging assumptions, or preparing a doc for executive review.
Validate the why before building through four product diagnostics — a YC-style product diagnostic that produces PRODUCT-BRIEF.md with a go/no-go recommendation, a founder review scoring product-market-fit signals, a user journey audit measuring time-to-value, and ICE feature prioritization. Use when pressure-testing product direction, choosing between features, sanity-checking a launch, or converting a vague idea into a product brief.
You MUST invoke this skill before saying “I don't know,” guessing, or treating any topic as new, no matter how trivial the question seems. It supplements other memory systems, which only hold partial records. Searching past conversations is the only way to recover what was actually said.
Reorganize the user's X and LinkedIn network with review-first pruning, add/follow recommendations, and channel-specific warm outreach drafted in the user's real voice. Use when the user wants to clean up following lists, grow toward current priorities, or rebalance a social graph around higher-signal relationships.
Operate execution flow across GitHub and Linear by triaging issues and pull requests, linking active work, and keeping GitHub public-facing while Linear remains the internal execution layer. Use when the user wants backlog control, PR triage, or GitHub-to-Linear coordination.
Use the Notion CLI (`ntn`) to interact with the Notion API, manage workers, and upload files. Use when the user asks to "call the Notion API", "deploy a worker", "upload a file to Notion", "create a page", "query a database", or any task involving the `ntn` command.
Plans a developer-platform career from the candidate side — API Product Manager, Platform Engineer/Architect, Partner/Integration Engineer. First clarifies "platform engineer": elsewhere in the industry, the term overwhelmingly means internal developer platforms (Kubernetes, Backstage, golden paths); this skill covers only the external/public-API-platform track. States the key finding that distinguishes this track from DevRel: platform compensation follows the general engineering/PM ladder, not a separate, under-surveyed one. Covers GitLab's real Partner Engineer ladder, the API-design-vs-system-design interview split, and RFC/design-doc portfolio signals. Use when the user asks how to break into platform/API product work, prepare for a platform or API-PM interview, or evaluate an offer. Do NOT use for hiring (samber/developer-platform-skills@developer-platform-hiring) or advocate/community/educator roles (samber/developer-relations-skills@devrel-career).
Assists in writing high-quality content by conducting research, adding citations, improving hooks, iterating on outlines, and providing real-time feedback on each section. Transforms your writing process from solo effort to collaborative partnership.
Route and complete Composio work across Composio For You and Composio Platform. Use when the user mentions Composio; wants an agent to use apps such as Gmail, Slack, GitHub, Notion, Calendar, or Linear; needs first-time setup, an SDK or MCP integration, CLI operation, migration guidance, current documentation, or help diagnosing a connection or tool call.
Use when building or maintaining a personal LLM-powered knowledge base. Triggers: ingesting sources into a wiki, querying wiki knowledge, linting wiki quality, 'add to wiki', 'what do I know about', or any mention of 'LLM wiki' or 'Karpathy wiki'.
WeCom (企业微信) Drive (Disk) file operation skill. Handles listing, searching, reading metadata, uploading, downloading, renaming and creating folders for files in 微盘 (Drive). You must read this skill first for full instructions whenever the user explicitly mentions "微盘", "网盘" or "共享空间" — do not act from memory. Use this skill when the user says "upload to 微盘", "search for xxx in 微盘", "where is that PPT in 微盘", "download that file from 微盘", "rename that 微盘 file to xxx", or provides a 微盘 file link in the form `https://drive.weixin.qq.com/s?k=...`. Difference from `wecomcli-doc` / `wecomcli-sheet` / `wecomcli-smartsheet` / `wecomcli-smartpage`: this skill handles file-level operations on all files in 微盘 (including online documents) — search, listing, basic info, location, path, renaming; reading and writing the contents of online documents (types `doc/sheet/smartsheet/smartpage`) belongs to the corresponding document skills and is not handled here. When the user asks "where in 微盘 is this document" or about a file's location in 微盘, this skill returns metadata such as space name, folder name and path via get.
WeCom Work calendar management. Triggers when the user needs to schedule an event, book a meeting room, view/update/cancel an event, or check free/busy time. This skill handles "scheduled events", i.e. arrangements without an online meeting link (this also covers purely offline face-to-face meetups); if the user wants an "online meeting" (with a meeting number/join link, remote or video participation), use the wecomcli-meeting skill instead. If the user only says something like "have a meeting / set up a meeting / that meeting" without specifying whether to create a scheduled event or an online meeting, you must first read this skill and ask the user to confirm using its disambiguation flow before acting — never create it by assumption.
WeCom email: send/reply/forward emails, search the email list, get email details (body, attachments, inline image parsing), and support sending schedule invitations and booking meetings via email. Use when the user needs internal email sending and receiving, email search, email management, etc. Note: schedules and meetings have separate skills; use this skill to handle meeting and schedule emails only when the user explicitly mentions "mailbox" or "email" (e.g., "send a meeting invitation via email", "send a meeting email").
WeCom Work media file upload/download skill. It covers two operations: downloading a media file to a local path based on media_id, and uploading a local file to obtain a media_id. When another skill (WeDrive, mail, etc.) returns a media_id that needs to be saved as a local file, or when an existing local file needs to be turned into a media_id for another skill, you must first read this skill to get the complete guidance — never work from memory. This skill does not parse or recognize file content; it only handles transferring the files themselves.
Query the range of chat sessions that can currently send messages, and send text, Markdown, images, files, voice, or video messages to individual or group chats in the session list. Use when the user asks to "message someone", "notify a group", "message the most recent chat", or "send an image/file/voice/video to WeCom".
Web search and content extraction with Tavily and Exa via inference.sh CLI. Apps: Tavily Search, Tavily Extract, Exa Search, Exa Answer, Exa Extract. Capabilities: AI-powered search, content extraction, direct answers, research. Use for: research, RAG pipelines, fact-checking, content aggregation, agents. Triggers: web search, tavily, exa, search api, content extraction, research, internet search, ai search, search assistant, web scraping, rag, perplexity alternative
Use this skill to query your Google NotebookLM notebooks directly from Claude Code for source-grounded, citation-backed answers from Gemini. Browser automation, library management, persistent auth. Drastically reduced hallucinations through document-only responses.
Universal deep research agent team. 13-agent pipeline for rigorous academic research on any topic. 8 modes: full research, quick brief, paper review, lit-review, fact-check, three-way literature scan, Socratic guided research dialogue, and systematic review with optional meta-analysis. Covers research question formulation, Socratic mentoring, methodology design, systematic literature search, source verification, cross-source synthesis, risk of bias assessment, meta-analysis, APA 7.0 report compilation, editorial review, devil's advocate challenges, ethics review, and post-research literature monitoring. Triggers on: research, deep research, literature review, systematic review, meta-analysis, PRISMA, evidence synthesis, fact-check, WHY HOW WHAT papers, 3W literature scan, guide my research, help me think through, 研究, 深度研究, 文獻回顧, 文獻探討, 系統性回顧, 後設分析, 事實查核, 三段式文獻掃描, 引導我的研究, 幫我釐清, 幫我想想, 我不確定要研究什麼, 研究方向, 研究主題, 심층 연구, 문헌 조사, 체계적 문헌고찰, 메타분석, 사실 확인, 연구 방향을 잡아줘, 연구 주제 정하는 것을 도와줘, revisión de literatura, metaanálisis
Search the web via the Bright Data CLI — `bdata search` for Google/Bing/Yandex SERP, `bdata discover` for intent-ranked semantic results. Use when the user wants SERP results, needs URLs to feed into scraping, or wants semantic web discovery with optional page content. Hands off to `scrape` once target URLs are chosen, and to `data-feeds` when the user wants structured data from a known platform. Requires the Bright Data CLI; proactively guides install + login if missing.
Two-tier memory system that makes Claude a true workplace collaborator. Decodes shorthand, acronyms, nicknames, and internal language so Claude understands requests like a colleague would. CLAUDE.md for working memory, memory/ directory for the full knowledge base.
View and edit cmux settings in ~/.config/cmux/cmux.json, including terminal, browser, Markdown, diff, notes, HTML, file preview, and sidebar-tool behavior. Use when the user wants to change cmux preferences, set a value by JSON path, validate the file, open it in an editor, or look up which keys cmux recognizes. Triggers on '/cmux-settings', 'change cmux setting', 'customize viewer', 'set <something> in cmux', 'cmux config', 'cmux.json', or 'rebind a cmux shortcut'.
Orchestrator for the full academic research pipeline: research -> write -> integrity check -> review -> revise -> re-review -> re-revise -> final integrity check -> finalize. Coordinates deep-research, academic-paper, and academic-paper-reviewer into a seamless 10-stage workflow with mandatory, coverage-bounded integrity checks, two-stage peer review, and auditable quality-assurance artifacts. Triggers on: academic pipeline, research to paper, full paper workflow, paper pipeline, end-to-end paper, research-to-publication, complete paper workflow, 연구부터 논문까지, 연구 주제 설정부터 논문 완성까지, 논문 전체 워크플로, flujo de trabajo académico, investigación a artículo, flujo completo de artículo, pipeline de investigación completa, publicación de investigación, flujo de trabajo completo del artículo.
Any live-web task via the Firecrawl CLI — including ordinary web research: searching the web, reading or extracting pages, gathering sources, discovering site URLs, bulk extraction, downloading a site, change alerts, or pages needing clicks/login — web only; local files route to firecrawl-parse. For US legal or regulatory questions use gov; for papers use firecrawl-research-index; for library, API, error, or bug questions use firecrawl-developer-index.
Operate on Feishu Base: create tables, fields, records, and views; handle statistics, formulas/lookups, forms, dashboards, app mode (BaseApp/AppMode pages and components), Workspace directories, workflows, role permissions, and the template center (Base template categories, lists, and search). Use when encountering Base/multidimensional tables/bitable, BaseApp/AppMode, or /base/ or /app/ links. Do not use lark-apps for BaseApp; route file import/export to lark-drive and authentication/authorization to lark-shared.
Feishu cloud storage (Drive): manage Drive files and folders, including uploads/downloads, folder creation, copying/moving/deleting, viewing metadata, checking permission settings, comments/permissions/subscriptions, titles, versions, Feishu document security labels (secure labels), and local file imports. Use when users need to organize Drive folders, handle cloud storage resource URLs/tokens, identify link types/actual tokens/titles, or import Word/Markdown/Excel/CSV/PPTX/.base as docx/sheet/bitable/slides. Route doubao.com cloud storage URLs/tokens by resource path and token too; do not fall back to WebFetch. Not responsible for: editing document content (use lark-doc), spreadsheet/Base table data operations (use lark-sheets/lark-base), knowledge space node/member management (use lark-wiki), or native Markdown file read/write/patch/diff (use lark-markdown).
Feishu Calendar: manage calendar events and meeting rooms. View/search events, create/update events, manage attendees, check availability and recommended time slots, and book meeting rooms. Use when users need to view schedules, create/modify meetings, or check/book meeting rooms. Not responsible for searching past video meeting records (use lark-meeting) or to-do tasks (use lark-task).
Feishu instant messaging: send and receive messages and manage group chats. Send and reply to messages, search chat history, manage group members, upload and download images and files, manage emoji reactions, send urgent notifications in-app/by SMS/by phone, send and handle Interactive Cards, and listen for card button callbacks (card.action.trigger). Use when users need to send messages, view or search chat history, download chat files, view group members, search groups, create group or topic chats, manage marker data, manage Feed pins (add/remove/search pinned conversations), manage tag data, or handle card callbacks.
Lark Approval: query and handle approval pending/completed/instances, search approvable definitions, view definition details and initiate native approval instances. Use when users need to handle approval tasks, view approval instances, search or initiate approvals. Approval pending items are not Lark tasks; non-approval pending items use lark-task. Not responsible for creating approval definitions; third-party approval definitions do not use native submission.
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.