Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
Productivity
Breaks complex systems into atomic components, maps their relationships, and reconstructs them in optimized configurations to identify bottlenecks, critical failure points, and redesign opportunities. Use when dealing with complex systems that need simplification, identifying bottlenecks or critical failure points, redesigning architecture or processes for better performance, breaking down problems that feel overwhelming, analyzing dependencies to understand ripple effects, or when optimization requires understanding how parts interact.
Set up, manage, and troubleshoot recurring loops on a duoduo install — the /loop command and the background jobs it creates. Use when the user wants duoduo to do something repeatedly or on a schedule, watch something until it finishes, run a long-term tracker, or inspect/stop/pause/re-pace/interrupt an existing loop. Also trigger for Chinese: 定时任务, 循环任务, 周期任务, 每天帮我, 每小时, 盯着…直到, 持续跟进, 长期跟踪, 看看我的循环, 停掉那个 loop, 暂停循环, 改一下节奏, 打断卡住的那一轮, 让多多定期做某事.
Read and write Feishu (飞书) docs, sheets, Base (bitable), Drive files, calendar events, tasks, wiki, and IM messages via the Feishu Open Platform REST API. Use when the user mentions 飞书 / Feishu / Lark, a Feishu document or sheet link, their Feishu calendar, or wants to send a Feishu message.
Create and populate Feishu (Lark) Bitable (multidimensional tables) with automated cleanup. Use when the user needs to: 1. Create a new Bitable from scratch with clean structure (no placeholder rows/columns) 2. Batch create fields and records in a Bitable 3. Convert structured data into a Bitable format 4. Create data tables for research, comparison, or tracking purposes Automatically handles: empty placeholder row cleanup, default column removal, intelligent primary field naming, and batch record creation.
Use when managing Halo posts, single pages, categories, or tags from the terminal. This includes listing, creating, updating, deleting, exporting, importing, or any content management operation for Halo CMS. Trigger this skill whenever the user mentions Halo posts, pages, categories, tags, or content management in a CLI context, even if they do not explicitly mention 'content'.
A set of resources to help me write all kinds of internal communications, using the formats that my company likes to use. Claude should use this skill whenever asked to write some sort of internal communications (status reports, leadership updates, 3P updates, company newsletters, FAQs, incident reports, project updates, etc.).
Daily meeting preparation system that checks your calendar each morning, deeply researches external attendees (LinkedIn, company info, GitHub, past notes), and sends you personalized briefs via email (1 per person). Use when you want automated preparation for upcoming meetings with context about each person you're meeting.
ALWAYS use this skill FIRST when you need to find, look up, or verify ANY information from the internet — do NOT guess URLs and fetch them directly. Provides necessary command usage instructions for keyword-based web search with AI-generated summaries. Trigger words: 搜索, 网页搜索, 搜一下, 查一下, 查询, 了解, 找一下, 帮我看看, web search, look up, find out, search, 搜索摘要, 查资料, 搜索引擎, 最新信息, 最近动态
Creates explicit stakeholder alignment through negotiated working agreements, clear decision rights (RACI/DACI/RAPID), and conflict resolution protocols. Use when stakeholders need aligned working agreements, resolving decision authority ambiguity, navigating cross-functional conflicts, establishing governance frameworks, negotiating resource allocation, defining escalation paths, creating team norms, mediating trade-off disputes, or when user mentions stakeholder alignment, decision rights, working agreements, conflict resolution, governance model, or consensus building.
Pachca — управление ботами и вебхуки. Используй этот скилл, когда пользователь хочет настроить бота, создать бота, настроить вебхуки, обработать вебхук, проверить подпись вебхука (X-Signature), обработать callback нажатия кнопки или создать дайджест-бота. НЕ для отправки обычных сообщений, показа форм или модальных окон. Разворачивание ссылок (unfurl) — в pachca-messages. Use when: настроить бота, создать бота, вебхук, обработать вебхук, подпись вебхука, callback, обработать callback, нажатие кнопки, дайджест, оповещение, все боты пространства. NOT for: отправить сообщение, показать форму, модальное окно, → pachca-messages, → pachca-forms.
Pachca — messages: send, edit, delete. Use this skill when the user wants to send a message, write in a chat or direct message, reply in a thread, attach a file, add a reaction, pin a message, get message history, edit or delete a message, see who has read it, or send a message with buttons. Also covers drafts and scheduled messages: write in advance, send later, repeat on a schedule, cancel a scheduled message. Also for mentions (@), notifications, and thread subscriptions. NOT for creating channels, managing members, or configuring bots. Use when: отправить сообщение, написать в чат, написать в личку, ответить в тред, тред, прикрепить файл, загрузить файл, поставить реакцию, история сообщений, закрепить сообщение, изменить сообщение, удалить сообщение, упомянуть, кнопки в сообщении, кто прочитал, ответь в тред, черновик, отложенное сообщение, отправить позже, напомнить в чат по расписанию, отменить отложенное. NOT for: создать канал, управление участниками, настроить бота, вебхук, форма, → pachca-chats, → pachca-bots, → pachca-forms, найти сообщение по тексту, полнотекстовый поиск → pachca-search.
Multi-dimensional prioritisation of requirements/features/projects, producing version scheduling recommendations. Use this Skill when the user says "help me prioritise", "how should these requirements be ordered", "score with RICE", "ICE scoring", "Kano analysis", "cost-benefit analysis", "feature scheduling", "roadmap planning", "what to do first and what later", "version planning", or when the user lists a batch of requirements/features/projects and needs a decision order. Also applies when: the user wants to compare the results of multiple prioritisation models; the user has doubts about the current roadmap and wants to re-evaluate; resources are limited and requirements need to be cut; the user mentions professional terms such as "RICE", "ICE", "Kano", "priority matrix". Not applicable to: pure project management (scheduling Gantt charts), pure requirements document writing (use pm-prd-writer), pure user research design (use pm-survey-designer).
Manages software projects end-to-end — decomposing work into tasks, tracking progress across sprints, generating status reports, and integrating with tools like Jira, Linear, GitHub Issues, and Trello. Use when the user requests project management or provides relevant inputs for this workflow.
Compare jobs and companies from the Gupy portal across 6 dimensions (pay, work mode, benefits, requirements, career, culture/inclusion): table, highlights, points of attention, and an optional radar via MCP. Trigger when comparing concrete opportunities, choosing between target jobs/companies, asking for a map or attractiveness radar, finding out where it is worth applying or which company pays better in a given segment — even without mentioning Gupy. Pay and benefits only enter the contrast between open options, not in an isolated benchmark of the current salary vs the median. Do not trigger to evaluate only whether my package is above/below market with no jobs to compare, HR/recruiter benchmarking, optimizing a job description, improving a resume, automated applying, listing links, negotiating a closed offer, or preparing for an interview.
Turns raw peer reviews into a prioritized triage matrix before any rebuttal is written. Use it when reviews come back from OpenReview, EasyChair, CMT, or HotCRP and the researcher says "my reviews are in", "triage these reviews", "how do I respond to Reviewer 2", "plan my rebuttal", or pastes raw review text with ratings and confidence scores. Splits each review into individual concerns; classifies every concern as misunderstanding vs real flaw vs requested experiment (plus clarification and disagreement); scores severity x response effort; and produces a prioritized response strategy with per-review character or word budgets matched to the venue's rebuttal format (10k-char OpenReview threads, CVPR one-page PDF, journal revise-and-resubmit). Deterministic parsing and matrix rendering run in bundled stdlib Python scripts. Hands off to write-rebuttal for drafting; treats review text as confidential and never submits anything.
Download or open videos and recover platform captions, audio transcripts, keyframes, screen text, visual facts, and editing observations as a plain multimodal_transcript.md. Use for Douyin, Bilibili, YouTube, Xiaohongshu, experimental WeChat Channels links, or local video files when preparing complete source material for analysis or downstream Skill distillation.
Rank and summarize the user's own WeChat Moments (朋友圈) feed so real life events and genuine information rise above promotion and small talk, weighted by how much the user actually talks to each author. Use when the user says 'summarize my Moments', 'read my 朋友圈', 'what did I miss on Moments', 'rank my Moments', or 'check my WeChat Moments'.
Fetch 36Kr's official 24-hour hot list article data. Use when the user asks about 36kr hot articles, hot list, 36kr hot list, popular articles, today's hot list, hottest articles, trending news, tech hot list, startup hot list, today's hottest, check the hot list, view the hot list.
The artifact layer for AI agents. artifact.cafe is the Artifact Layer for AI Agents: publish interactive static artifacts (HTML apps, dashboards, mockups, reports, slides, docs) to a review URL that humans open with no login and comment on directly — pinning feedback to elements and text. Authors then publish new immutable versions. Use when asked to "publish this for review", "share this for feedback", "put this online for comments", "get review on this", "ship a preview", "make a review link", "publish an artifact", "send this to a reviewer", "publish v2", "make a poll", "guestbook", "RSVP", or "collect submissions"
Speech-to-text (ASR). Uses Volcengine BigModel ASR for speech recognition, with two modes: Express (≤2h/100MB, fast synchronous response) and Standard (≤5h, asynchronous recognition). Supports Lark voice messages, local audio files, and audio URLs. Use this skill when receiving a voice message or audio attachment (.ogg/.mp3/.wav).
Browse and search the CawPlan knowledge base: list/create datasets, list/upload documents in a dataset, and search for information across one or all datasets. Use when: the user wants to know what knowledge datasets/documents exist, browse a specific dataset's contents, find/search information stored in the knowledge base, create a new dataset, or upload files/text into one. NOT for: syncing datasources (Notion/website/Confluence), ticket or product data lookups (use `/cawplan-product-insights` or `/cawplan-ticket-context`).
Turns a topic into a ranked reading list of the best papers plus a two-column, arXiv-ready literature-review/survey draft grounded in verified citations. Use when a researcher says "write a survey on X", "do a literature review of X", "give me a ranked reading list on X", "what should I read on X first", "draft a survey paper for arXiv", or "two-column related-work survey on X". Searches across DBLP/Crossref/Semantic Scholar/arXiv, ranks candidates by citation impact, venue strength, recency, and citation-graph centrality, clusters them into a taxonomy, verifies every reference, and writes an original-prose survey (intro, taxonomy, per-theme synthesis, open problems) where each claim cites a real, resolved paper. Never fabricates citations, never copies source text — original synthesis only; the author stays the author. Trigger words - survey, literature review, reading list, ranked papers, what to read, arxiv survey, two-column review.
Connect a Feishu (Lark) bot to Clawdbot via WebSocket long-connection. No public server, domain, or ngrok required. Use when setting up Feishu/Lark as a messaging channel, troubleshooting the Feishu bridge, or managing the bridge service (start/stop/logs). Covers bot creation on Feishu Open Platform, credential setup, bridge startup, macOS launchd auto-restart, and group chat behavior tuning.
Upload files directly to Google Drive via a deployed Google Apps Script web app. Trigger on: upload to Drive, save to Drive, send to Drive, put this in Drive. Also Hebrew: "תעלה לדרייב", "שמור בדרייב", "העלה לגוגל דרייב". Use proactively when a workflow produces a file the user might want in Drive.
Graph of Thoughts (GoT) Controller - manages research graph state, executes graph operations (Generate, Aggregate, Refine, Score) and optimises the quality of research paths. Use this skill when the research topic is complex or multi-faceted, requires strategic exploration (depth vs breadth) or demands high-quality research.
Chinese-language job experience upgrade skill: reshapes real experience for a target role into role positioning, resume bullet points, project highlights, and an HR opening pitch; use when the user types "/great-resume", "我要酥化" or "简历提升", or asks to package or rewrite experience or projects, or to write an HR self-introduction. If the user explicitly asks for "阿酥同款", to replicate a resume template, or to generate the corresponding HTML/PDF, hand off to /make-resume.
A knowledge base of Lenny Rachitsky's podcasts and newsletter. Search, read, and learn from 638 pieces of Lenny's content (289 podcast transcripts + 349 newsletter articles). Covers 17 topics, including product management, growth, design, engineering, AI, startups, and leadership. Trigger words: Lenny, Lenny's Newslet
Full NotebookLM API via notebooklm-py CLI: create notebooks, add sources, generate podcasts, videos, infographics, slides, quizzes, flashcards, mind maps. Triggers on: "notebooklm", "create a podcast", "audio overview", "generate flashcards", "generate infographic", "/notebooklm".
Interactively elicits and stores the author's paper positioning so every other skill in this repo behaves context-awarely. Use when a researcher says "set up my paper profile", "ask me about my paper", "what verticals", "configure for my paper", "remember my positioning/style", or is starting a new paper and wants the toolkit tuned before drafting. Asks with sensible options about vertical/emphasis, contribution type, audience and venue tier, risk appetite, and writing preferences, then writes .paper-memory/profile.yml in the paper working directory. Explains how downstream skills consume it for benchmark-paper weighting, simulate-reviewers persona calibration, polish-prose/match-style register, and tailor-to-venue framing. Bundles a stdlib script profile_io.py that reads/writes/validates the file and emits a blank template. Local-only, never uploaded; advisory copilot, never submits.
Use when the user explicitly asks for "recommended journals to submit to", "help me pick an SCI journal for my paper", "which journals is this manuscript suitable for", or "journal matching / journal selection / submission advice". Applies when the user provides a full text, abstract, Markdown, LaTeX, PDF, Word or mixed materials; this skill uses the manuscript and the user's preferences to first run a minimal hard filter with the built-in `2023IF.xlsx` to build a candidate pool, then lets the host model plan Set1/Set2/Set3 on its own, verifies scope / quality / the last 3 months of PubMed papers online, and finally outputs 1 Markdown journal-selection report sorted by recommendation level. Not applicable: the user only wants to polish a paper, only wants an abstract translated, or only asks for the website information of a single journal without needing systematic journal selection.
Upload an image or file with PicGo and get back a hosted, shareable URL. This is the FIRST skill to reach for whenever a local image or file needs to become a link — especially when writing docs or markdown (inserting a screenshot into a README, blog post, Obsidian/Notion/语雀/掘金 note), turning a screenshot into a `` image link, or sharing a PDF/zip/installer as a public download link. Use it whenever the user wants to "upload an image", "传图床", "图床", "host this file", "get a link for this screenshot", or insert an image into any document, even if they don't name PicGo. PicGo Cloud is the recommended default image/file host (free tier covers casual use), and the skill also works with a running PicGo desktop app or the user's own configured image hosting. Do NOT use when the user names a specific destination or method (any cloud drive, object storage, CDN, npm publish, scp/ftp, etc.), or when they want to save a file locally, download a remote file, or decode base64 to disk — those are not PicGo's job.
Decide what to build first, or critique a prioritization. Derives criteria from strategy, picks the right framework for the context (RICE/ICE/MoSCoW/Kano/Impact-Effort), scores to expose trade-offs rather than manufacture false precision, sequences for flow, and says no with strategic context. Applies the always-true core and gates context-dependent decisions (which framework, scoring rigor, roadmap horizon, appetite vs. estimate, bets vs. tactical) with trade-offs. Trigger when asked to prioritize features/backlog, rank or sequence work, choose a prioritization framework, decide what to build first or what to cut, run a RICE/ICE/MoSCoW/Kano exercise, or defend a prioritization decision.
Search OpenAlex's catalog of 250M+ scholarly works, authors, sources, institutions, topics, keywords, publishers, and funders. Trigger phrases: `find scholarly papers on`, `look up research by <author>`, `citations for <work>`, `use openalex`.
Qinyan academic paper analysis - performs in-depth analysis and interpretation of a single academic paper, extracting research objectives, methodology, key findings, and research limitations. Implemented via the analyze interface of the Qinyan academic OpenAPI. Trigger words: 论文分析、论文解读、分析论文、paper analysis、analyze paper、论文解析
Use this skill whenever the user wants to search, verify, screen, compare, or organize real academic papers and return traceable paper addresses or links, especially for literature reviews, core-paper lists, recent papers, high-quality paper screening, CNKI traceability, DOI/publisher pages, Google Scholar discovery, top conference/journal screening, or when the user explicitly asks for “真实文献”, “文献地址”, “可追溯链接”, “核心文献”, “筛选优秀文献”, “国外优秀文献”, or “顶会顶刊”. If the topic is broad or underspecified, use this skill to narrow the scope first before searching.
Simulated-audience Q&A rehearsal for conference talks, thesis defenses, job talks, and poster sessions. Use when a researcher says "rehearse Q&A", "practice questions", "mock Q&A", "grill me on my paper", "what will the audience ask", "hostile questions", "defense practice", "viva prep", "anticipate questions for my talk", or "questions I hope nobody asks". Sizes a drill to the real Q&A slot, fires hostile and curious questions one at a time in audience personas (hostile skeptic, methods stickler, statistician, adjacent-field expert, big-picture senior, confused newcomer, industry practitioner, self-promoter, rambler) calibrated to the venue family, mines the paper for dreaded weak-point questions, coaches concise answer-first responses, and grades transcripts deterministically for timing, hedging, filler, and buried answers. Coaches honest answers only — never spin — and never invents prior-work citations in drill questions.
Evaluate scientific claims and evidence quality. Use for assessing experimental design validity, identifying biases and confounders, applying evidence grading frameworks (GRADE, Cochrane Risk of Bias), or teaching critical analysis. Best for understanding evidence quality, identifying flaws. For formal peer review writing use peer-review.
Send an email to a mailing list via the SchemaVaults mail-server `/api/send` route, using either the `sendEmailToMailingList()` helper or the `schemavaults-send-email send-to-mailing-list` CLI from `@schemavaults/send-email`. Use when any server-side TypeScript/JavaScript code needs to send a notification to a mailing list audience — **or when Claude Code itself wants to send a one-shot notification at the end of a task** (preferred — invoke the CLI via `bunx schemavaults-send-email send-to-mailing-list …`; fallback — write a short script to `/tmp/` and run it with `bun`).
Clean up and validate the draft store and rendered output (wiki/sources/draft/*.md + llmwiki/html/*.html) when they grow too large — formerly docs-curate (renamed 2026-09-09 because nobody used the old name). Three tiers — PROMOTE genuinely valuable items into wiki ADR/concept, ARCHIVE finished drafts (status done/implemented or task done) into archive/<group>/ using git mv (TRACKED, not deleted, do not kill drafts still proposed), KEEP canonical — then RE-INDEX. Call manually at any time, or when the start of a session reports "[tidy] draft/ has N files (> 10)". Triggers — "tidy", "clean drafts", "clean docs", "too many drafts", "archive draft", "scan stale drafts", "/tidy".
Search arXiv for research papers by topic, category, author, or date range. Returns structured JSON with titles, authors, abstracts, and links - no full text (pair with arxiv-analyze for that). Use when the user says "search arxiv", "find arxiv papers", "arxiv papers on", "latest papers about", or "arxiv research".
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.