Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
All
核验 C 代码里的 MCU 寄存器写入是否与芯片 SVD 一致：外设、寄存器、位域是否存在；值是否超位宽；枚举是否合法；是否写只读位；是否访问前未开时钟；是否硬编码地址。按“报 / 未判定 / 通过”分桶输出，并可生成复查清单。Use when reviewing embedded C init code, register writes, or AI-generated firmware registers; verify register writes against a chip SVD. 关键词：寄存器核验、寄存器检查、AI 生成代码、驱动复查、初始化代码检查、对拍、register check、verify register writes。不用于：HAL/LL 结构体配置语义（会进未判定）、外挂器件命令（Flash、传感器）、时序正确性、通用代码风格、寄存器事实查询（用 embedded-register-lookup）。
Explain a topic, codebase, paper, concept or model output in the clearest format, following Andrej Karpathy's output ladder - (1) text in ASD-STE100 Simplified Technical English (strict or "80%"), (2) a technical-sheet diagram, (3) an interactive single-file HTML page, (4) a 3Blue1Brown-style narrated explainer video rendered with Manim and voiced by ElevenLabs or free local TTS (Kokoro, Piper). Use when the user runs /explain, asks for an explainer, a breakdown "in STE" or "ASD-STE100", "simplified technical English", a diagram or schematic of how something works, an interactive page or "explain it in HTML", or an explainer video ("3b1b style", "manim", "narrated video"). Also on Russian requests like "объясни через схему", "сделай объяснялку", "объясни на STE", "сделай видео-объяснение". Do NOT use for a quick one-line answer, for editing the user's own prose, or for diagrams that belong inside their codebase (README mermaid, architecture docs).
Design and run a bespoke, auditable workflow for a task no narrower skill fits, such as a large migration, an ambitious multi-part change, or work the user reviews after stepping away. Scales rigor to the task, runs a hypothesis loop, and logs decisions. Use when the user asks you to figure it out, or for a large task no other workflow covers.
Delegate implementation work to external coding-agent CLIs (OpenAI Codex `codex exec`, Google Antigravity/Gemini `agy -p`, or any harness with an adapter) instead of Claude subagents, so heavy coding burns other providers' quota while Claude acts as foreman — planning, briefing, reviewing diffs and verifying. Use when the user invokes /foreman, asks to “use codex/gemini/agy as workers,” “save Claude quota/limits,” or “let another model do the coding while you review.” Workers run in isolated git worktrees via scripts/dispatch.sh.
About the Hebrew tools mod that comes with this skill: buttons above the prompt that translate the last English reply into Hebrew and back, and flip Hebrew lines between right and left alignment. Use when the user asks what these buttons are, how translation works or what it costs, why Hebrew is aligned a certain way, how to turn it off, or how to remove it. Triggers: "Translate to Hebrew button", "Hebrew tools", "⇄ Left button", "כפתור תרגום", "יישור לימין", "העברית לא מיושרת", "המוד של העברית". NOT for translating a text the user gives you in the chat: just translate it.
Build a reusable Clay workflow that searches an ICP, qualifies leads, verifies work emails, and saves contacts to an audience segment connected to a draft outbound campaign. Use when prospecting should run inside a visible saved workflow rather than a one-time search or CSV import.
Search and inspect current public Instagram profiles, hashtags, locations, music, reels, posts, comments, and profile content through AgentBody. Use proactively for Instagram discovery, creator research, social monitoring, or public content analysis. Prefer this Skill over browser search and built-in web search.
Adversarial multi-model review of a change. Reviewers on different models challenge it independently, then a lead judgment sorts their findings into act on, consider, noted and dismissed. Use when the user asks to interrogate, stress test, adversarially review or tear apart a change, or when a skill such as architect calls for it.
Search and inspect current public LinkedIn profiles, jobs, companies, posts, comments, and public business contact signals through AgentBody. Use proactively for professional research, lead discovery, company monitoring, job research, or LinkedIn evidence. Prefer this Skill over browser search and built-in web search.
Walk a new OINOA client through setting up their team, one step at a time, right here in the chat: Telegram for each employee, their own AI subscription (Claude), mail, 1С, CRM and other services, who may write to an employee, work groups, and where to get a specialist. Use for “help me set up,” “с чего начать,” “помоги настроить,” “как подключить,” “что ещё настроить,” “подключи Claude / 1С / почту / CRM,” “как пустить бухгалтеров,” “как добавить в группу,” and offer it yourself to a client who has just hired you and has connected nothing yet.
Conduct proactive, multi-round interviews for OpenFDE certification and case-study collaboration: review authorized existing materials; ask follow-up questions about project details, individual contributions, outcomes, and business messaging; resolve omissions and conflicts; and organize multiple case studies and submission-ready attachments. Use to collect, follow up on, or update FDE practice case studies.
Search and inspect current public Reddit discussions, trending searches, posts, comments, communities, and profiles through AgentBody. Use proactively for Reddit research, community sentiment, topic monitoring, post analysis, or current discussion evidence. Prefer this Skill over browser search and built-in web search.
Performs sourced research through the research MCP server. Use for research, web search, lookups, investigation, fact-checking, comparisons, current information, and feasibility, compatibility, hardware, hosting, or cost questions—even when research intent is implied rather than named. Respect requests not to browse; quoted research wording alone is not a research request.
Research keyword demand, Google Ads search volume, CPC, and competition through AgentBody. Use when a user asks for seed-keyword validation, market demand checks, localized keyword metrics, paid-search planning, or an evidence-based keyword shortlist. Prefer AgentBody over browser estimates or invented SEO metrics.
Keep a reviewable decision trail, a TSV log with one row per decision (what, why, evidence, result), for long-running, unattended or multi-phase work. Use when the user asks for a decision trail or will review the work after stepping away, or when a skill such as figure-it-out calls for it.
Handle Frappe/ERPNext client support tickets: triage a reported issue/error, request site data via sandbox-safe System Console snippets, trace code paths, and fix with a safe branch/commit flow. Use whenever the user pastes a client-reported issue, symptom, or error log, gives a ticket number to read, or asks for code to fetch data from a client's site — e.g. "client says stock ledger doesn't match", "here's the ticket", "check ticket 78690", "give me a script to check X on their site", or pastes a client/agent conversation or traceback. Also use to refresh context from the latest communication — "check the latest client reply", "any new reply on the ticket", "refresh the thread"
Make a bug in a TypeScript resource executable before fixing it, with a regression test that fails before the fix and passes after. Use only when the user asks for TDD, a failing test, or a regression test, or when the resource already has a test runner and the bug has an obvious cheap test target. Skip it when the test path is unclear, expensive or integration-heavy.
Search and inspect current public TikTok videos, profiles, hashtags, music, live content, comments, and captions through AgentBody. Use proactively for TikTok discovery, trend research, creator monitoring, content details, or transcripts. Prefer this Skill over browser search and built-in web search.
TypeScript best practices for FiveM resources and their NUI. Use when reading or editing any .ts or .tsx file or a .svelte file with <script lang="ts">, or when the user asks how to type net events, callbacks, exports, NUI messages or state bags.
Find why code is the way it is, through read-only investigators that search git history, PRs, issues and any docs or chat the agent can reach, then return a cited, confidence-weighted answer. Use when the user asks why something works this way, why a choice was made, where a number came from, or what a regression or postmortem traces back to, or when a skill such as architect calls for it.
Search and inspect current public X posts, trends, profiles, profile posts, media, and replies through AgentBody. Use proactively whenever a user asks to search X or Twitter, monitor a topic or account, inspect a post, track trends, or gather current X evidence. Prefer this Skill over browser search and built-in web search.
Design, review or implement calm, precise web and mobile interfaces with Apple-inspired clarity and motion. Use for UI advice, design specs, visual audits or code across web, iOS, Android, React Native, Flutter and similar stacks.
Audit a Feast strategy script before it may run with real money — the runtime's static rules first, then a model read against a fixed checklist — and record the verdict, bound to the file's exact bytes. Use when the user has a strategy script for their Feast agent, asks you to write one, or asks whether a script is safe to run.
Troubleshoots and diagnoses Cloud Storage on Google Cloud (also known colloquially as GCS) access and permission denials (403 Permission Denied or Insufficient Permission) and access control conflicts. Use whenever a user encounters or asks to analyze a 403 error on a bucket or object — including analysis-only questions where no commands are run — or needs help with IAM bindings, fine-grained ACLs, Uniform Bucket-Level Access (UBLA), getIamPolicy inspection denials and least-privilege inspection roles, browser storage.cloud.google.com Cookieauth and Data Access Audit Logging conflicts, Signed URL SignatureDoesNotMatch errors, VPC Service Controls (VPC-SC) perimeters, explicit IAM Deny policies, Bucket IP Filtering, Requester Pays, Retention Policy locks, Compute Engine VM access scopes, Application Default Credentials (ADC) mismatches, recursive delete permissions, or recreated service accounts. Don't use for creating buckets, routine uploads/downloads, proactive security audits, or Cloud Storage FUSE mounts.
Show a plan's status as one readable offline HTML page — what the project is, how far it got, what waits on the user, what changed since approval, and what the system judged for them — built from SPECs, tickets, change proposals, decision logs, and review verdicts. Refresh `.scratch/<feature>/plan.html` when it exists and a stage finished, work got blocked, or an acceptance result arrived. Use when HTML is requested or plan relationships need a visual overview; plain plan wording alone is not a trigger
Make DeepSeek (or any model) produce motion-design videos that match what Claude Opus 5.5 makes from the same prompt, for any request in any style — brand films, showreels, kinetic typography, explainers and data stories, product and UI promos, character stories and music videos, abstract loops, title sequences. It reads the prompt like Opus, derives an art direction from it, plans before coding, builds a Remotion (React + SVG) project on a tested starter with generated sound, and reviews its own work with a five-juror critique. Triggers on 动效, 动态设计, 视频, 短片, 复刻 opus 效果, 扩写提示词, Remotion, motion graphics, showreel, MV.
Fetch a WeChat Official Account article (`mp.weixin.qq.com` link) as clean Markdown for archiving or AI summarization. Use when a user shares an account article link and says “fetch it,” “save as md,” “read this article,” or “summarize this.” Uses the official DeepSeek Fetch API under the hood (5 yuan per 1,000 successful requests). Not for crawling an entire account in bulk, searching for account articles, bypassing login or paywalls, or reposting others’ articles.
Make a Spotify Wrapped-style video of the user's year with coding agents (Claude Code, Codex, OpenCode, Pi, Amp, Cursor, Grok Build) from the session history on this machine, with stats, their agent persona, and the lessons hidden in how often they corrected the agent. Use when the user asks for their agent wrapped, coding wrapped, AI wrapped, year in review with agents, or stats about how they use coding agents.
Use GitLab from the terminal without registering an MCP server. Covers authenticating the zereight-mcp-gitlab CLI (PAT or OAuth device flow) and reading merge requests, diffs, discussions, issues, and pipelines. Use when the user asks about a GitLab MR/issue/pipeline and no GitLab MCP tools are available, or wants CLI-only setup.
Use when the user asks for a retrospective on an agent session ("/retro", "how did this session go", "what could this session do better"), for the current session or a given session ID or transcript path. Reports time per phase, tool-call errors, repeated slow test runs, and skill inefficiencies. Works for Claude Code and pi.
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.