Skills
100,696 packaged capabilities and playbooks for agents. Install any of them with one command.
All
Send an independent judge that is not the author: a fresh evaluator subagent that checks finished work against pass criteria written before it was built (evaluator mode), or audits a batch against its source of truth (audit mode), with evidence for every verdict and never a fix. Use when the owner asks for an evaluator, reviewer, auditor or judge, or says 'send a judge', 'evaluate this', 'check my work with a fresh agent', 'get a second agent to review this', 'audit this folder', 'audit my registry', 'verify this against the source of truth', or 'LLM-as-judge'. Runs only when asked; it does not fire on its own at the end of a task. Not for checks a machine can run (tests, schema validation, a search that must return zero), not for hundreds of identical units, and not for fixing anything.
Use when an agent needs live China travel data — geocoding, POI search, driving routes and map tiles (高德 Amap), flight schedules, transfers and fares (飞常准 Variflight), high-speed rail tickets (携程 Ctrip), weather and city coordinates (Open-Meteo) — or the planning order and mode trade-offs behind a trip. Triggers on 查航班 / 查火车票 / 查天气 / 查景点坐标 / 查驾车距离 / 行程该先定什么. Not for generating a full itinerary or shortlisting attractions.
Use this skill when the user asks questions about SAP Clean Core Extensibility — including extension levels (A–D), on-stack vs. side-by-side extensibility, ABAP Cloud development model, ATC setup, KPI measurement (clean core share, technical debt score, unused code share, business modifications), Governance & Maturity Assessment practices (EXT-GOV-01 through EXT-SYS-05), RISE with SAP Methodology, Project Kernseife, the SAP API Policy, SAP Build and Joule Studio, or Get Clean / Stay Clean patterns. Also use when the user asks about the RISE with SAP Methodology Dashboard Extensibility section — Technical Debt Score, Clean Core Level distribution (A/B/C/D), Execution of Customer Objects, Business Modifications, Unused Customer Objects, ATC level mapping, the Customer Objects deep-dive, the Business Modifications deep-dive, or Object Type Harmonization (effective August 2026). For the full dashboard reference, use `clean-core-general`. For broad clean core overview questions, use `clean-core-general` first.
Use this skill when the user asks broad or introductory questions about SAP Clean Core — including what clean core is, why it matters, the five principles (Extensibility, Integration, Data, Business Processes, Operations), the Autonomous Enterprise vision, RISE with SAP Methodology, general KPI measurement across all principles, the cross-principle Governance & Maturity Assessment, how to start a clean core journey, or how to explain clean core to stakeholders. Also use when the user asks about the RISE with SAP Methodology Dashboard in SAP Cloud ALM — what it is, what it measures, how to set it up, the System View, the Operations View, the deep-dive views, data collection tools, or dashboard FAQ topics. Also use when the question spans multiple principles. For deep questions on a single principle, use the principle-specific expert skill.
Use this skill when the user asks questions about SAP Clean Core Operations — including what clean core operations is, the four cornerstones of clean operations, Conway's Law and its impact on SAP transformations, the Customer Centre of Expertise (cCOE), Solution Standardization Board (SSB), Operations Control Center (OCC), SAP Activate and RISE with SAP Methodology, governance and operational practices (OPS-01 through OPS-08), the six Operations KPI categories, software stack currency, test management, software delivery efficiency, SAP Cloud ALM, SAP Focused Run, SAP Secure Operations Map, or the Autonomous Enterprise vision. Also use when the user asks about the RISE with SAP Methodology Dashboard Operations View — the Operations Score, Security score, System Health score, or Operations Setup score and their metrics, SLAs, and calculation formulas. Also use when the user asks about KPIs or governance maturity for the operations dimension. For broad clean core questions, use `clean-core-general` first.
Use this skill when the user asks questions about SAP Clean Core Business Processes — including what the business processes principle is, business process management (BPM), the fit-to-standard concept, SAP Activate methodology phases (Discover/Prepare/Explore/Realize/Deploy/Run), delta requirements, fit/gap classification (F-SC, F-VS, F-EC, G-EC, G-RP, G-EP, G-EN, G-ET, G-AN), the Solution Standardization Board (SSB), Key Decision Documents (KDDs), standard process archetypes, SAP content for fit to standard, the recommended tools (SAP LeanIX, SAP Signavio Process Transformation Suite, SAP Cloud ALM), the 4 process KPIs (Fit-to-Standard Alignment, FTS Alignment – Gaps, Alignment to Clean Core, Alignment to S/4HANA Target Architecture), or Governance & Maturity Assessment practices for Business Processes (PRO-01 through PRO-05). For broad clean core overview questions, use `clean-core-general` first. For the RISE Dashboard reference, use `clean-core-general`.
Finds, buys, or connects a custom domain (e.g. meinladen.ch) for the Albatros server. Explains domains and DNS with the analogy of an address and phone book, suggests names, checks availability with domain_suchen, and buys through Albatros using domain_kaufen (the person is the owner and enters their details only on the payment page). If the person already has a domain, the skill connects it with domain_verbinden: DNS records, TXT proof _albatros, and checks until the site appears with a lock in the browser. Use this skill for “custom domain,” “my address,” “buy .ch,” “which domain is available,” “search for a domain,” “DNS,” “A record,” “CNAME,” “TXT record,” “domain doesn’t point to my site,” “certificate,” and “https isn’t working.”
Turn stills into video and do media surgery from the shell — Ken Burns motion, crossfades, narration muxing, contact sheets, frame extraction, GIFs, loudness normalisation and speed changes. Use when the user wants a video, a slideshow, a preview clip, a contact sheet of a video, or needs audio and video combined, trimmed or normalised.
Group a scope of Issues into Waves that can be implemented concurrently without stepping on each other — estimate what each Issue touches, exclude the ones that need a human, and print launch lines. Use when the user has several prepared Issues and wants to know which can run in parallel unattended
Search ~1.01M computer-science papers and a curated research wiki with the `fs` command (FrontierScholar CLI). Use it whenever a task needs CS literature, even if FrontierScholar is not mentioned. This covers finding papers on a topic or before a date cut-off, building a related-work or reading list, finding papers similar to given ones, checking a paper's abstract, method, evaluation or first-version date, finding which papers use a benchmark or dataset, and finding the open problems and rising approaches of a research area. Prefer it over memory for paper ids, titles and dates.
Explains Git to someone with no prior knowledge using the analogy of a logbook aboard a ship and its copy in the harbor (Albatros's Git service, code.aufwind.sh). Gets the repo from the harbor, guides them through their first entries (Commit) with good notes, sends them to the harbor (Push), and checks that the copy arrives on the Albatros server at ~/projekt. Rules: main only, stage only your own files individually, never force-push. Use this skill for “Git,” “GitHub,” “harbor,” “commit,” “push,” “pull,” “Repo,” “what is a repository,” “save my changes,” “push rejected,” and “conflict,” and before committing anything for the person.
Helps someone with no prior knowledge write and choose text and images for their website: a homepage in three sentences, short text for people reading on mobile, their own photos instead of stock images, image rights and image size, and making AI text sound like their own voice. The result is a content list in docs/inhalte.md and the text on the site. Use this skill for “writing text,” “what should I write on the homepage,” “About us,” “images for the website,” “may I use this image,” “text sounds like AI,” and “content.”
Plans and guides a project's move from Albatros to its own server, explaining it with the analogy of moving into your own house: a moving checklist; renting and setting up a server (SSH with keys only, firewall, updates, Caddy for HTTPS); taking the Logbuch on GitHub along; switching Fliessband there to the new destination over SSH with your own key; bringing data and secrets; testing; changing the address in DNS, or transferring a domain bought through Albatros to a new registrar with the auth code from domain_mitnehmen; and handling backups yourself. Use this skill for “own server,” “move,” “away from Albatros,” “Hetzner,” “VPS,” “self-host,” “landing,” and “what do I need for my own server.”
Hauptskill von Albatros: der Fluglotse. Fragt mit dem Werkzeug lage den echten Stand der Person ab (Abo, Schlüssel, Server oder Seite, Fliessband, Domain, Mission) und folgt dann genau dem Skill, der jetzt dran ist; nach jedem geschafften Schritt fragt er neu. Ohne Anmeldung zeigt er die offene Bibliothek aller Skills zum Selbstwählen und erklärt, wie die Anmeldung geht. Nutze den Skill, sobald jemand an seinem Albatros-Projekt weiterarbeitet oder fragt «was jetzt?», «wie weiter?», «was ist der nächste Schritt?», «wo stehe ich?», «es geht nicht», «es klemmt», «ich weiss nicht weiter», bei /albatros:lotse und am Ende des Skills start. Bei einer klaren Frage zu einem Thema (z. B. «wie verbinde ich meine Domain?», «was ist ein SSH-Schlüssel?») hat der passende Fach-Skill Vorrang.
Use when the user wants to brainstorm, think an idea through, or explore what to build; brings a rough idea whose goal or reason is unclear (“add notifications”, “something for onboarding”, “I have an idea for an app”); or asks for ideas or options (“give me ideas for X”, “what could we do about Y”), in software or outside it. Works in Claude Code and Codex. Produces a user-approved design brief, ready for mstack-write-specs.
Use when the user wants Codex to write the code while Claude manages, reviews, and commits it, for a plan or plan folder, a list of issues (GitHub or another tracker), or one larger task. Examples: “let Codex develop this plan”, “Codex writes, you review”, “use the Codex plugin to implement these issues”, or when an mstack-plan execute-plan.md offers this workflow. Runs only in Claude Code with the OpenAI Codex plugin; for Claude subagents as writers, use mstack-codex-review-workflow. Produces one integration branch with one reviewed commit per phase.
Use when review feedback needs action, such as findings from a cross-vendor reviewer, advisor, fresh-context reviewer, CI review bots, or pull request comments; when the user says “process this review”, “address the review comments”, or “handle these findings”, pastes reviewer output, or another mstack skill needs a review processed. Works in Claude Code and Codex. Produces verified fixes with tests, reasoned rejections, and a log with a drafted reply for each finding.
Set up a "chief of staff" workflow on herdr. One chief agent routes requests to a manager agent per project, and managers run workers in git worktrees and herdr tabs, all in parallel. Use when the user asks to set up numbrone, a chief of staff or multi-project agent workflow on herdr, or to add a project to an existing numbrone setup.
Erklärt CI/CD einer Person ohne Vorwissen mit dem Bild vom Fliessband mit Qualitätskontrolle und richtet es ein: Ein Workflow im Hafen von Albatros (Forgejo, code.aufwind.sh) startet bei jedem Push auf main. Der Arbeiter (Runner) auf dem eigenen Server baut die Seite, liefert sie mit Gesundheitscheck aus, prüft sie über die öffentliche Adresse und springt sonst auf die letzte gute Version zurück. Nutze den Skill bei «Pipeline», «CI/CD», «GitHub Actions», «Forgejo Actions», «Workflow», «Runner», «automatisch ausliefern», «deployen», «wie kommt meine Änderung auf den Server», «Fliessband ist rot», «Deploy fehlgeschlagen».
Take over or hand off a Git code project without needing to re-explain it. Use this skill when the user wants to continue a previous project, switch computers and continue, hand the project to someone else or another agent, or mentions relay, picking up, continuing, handoff, switching computers or devices, another computer, or setting up relay—even without mentioning relay or project memory. At the start of a session, reconcile; at the end, collect. Only project goals need human confirmation; everything else is automated.
Helps bring a small Swiss website into legal compliance, explaining it to people with no prior knowledge using the analogy of a doorbell nameplate and house rules: legal notice (UWG Art. 3 Abs. 1 lit. s), privacy policy under revDSG (information duty Art. 19), cookies (Art. 45c FMG), when the EU GDPR also applies, image rights, and marketing emails only with consent. Not legal advice. Use this skill for “legal notice,” “privacy,” “privacy policy,” “DSG,” “DSGVO,” “cookie banner,” “do I legally need this,” “what must go on my website,” and “contact form privacy.”
Helps a small website become easier to find in search engines, explaining it to people with no prior knowledge using the analogy of signposts and a city map. Covers page titles and descriptions, a main heading, descriptive URLs, image descriptions, loading time, sitemap.xml, robots.txt, structured data for local businesses, Google Search Console and Business Profile. Use this skill for “SEO,” “get found on Google,” “Google can’t find my site,” “search engine,” “Sitemap,” “meta description,” “get found locally,” and “Google Maps listing.”
Rundgang durch den eigenen Albatros-Server für eine Person ohne Vorwissen. Erklärt Terminal, Ordner, Speicher, Arbeitsspeicher, laufende Programme und den Hauptschlüssel sudo mit Bildern aus einer Wohnung. Die Person tippt die ersten Befehle selbst. Nutze den Skill bei «was ist auf meinem Server», «wie bewege ich mich im Terminal», «was ist ein Ordner auf dem Server», «ls cd pwd», «wie viel Platz habe ich», «was läuft da», und nach dem ersten erfolgreichen SSH-Besuch.
Pick the right skill for an intent and print a pasteable command. Reads MANIFEST.json (or every SKILL.md frontmatter when there is no manifest) and scores the intent against use_when phrases, curated keywords and not_for exclusions. Use when an agent has several skills installed and must choose one entry point before doing anything, or when you want to know which script handles "make me a logo" versus "read this page".
Erklärt einer Person ohne Vorwissen den SSH-Schlüssel mit dem Bild von Hausschlüssel und Schloss, erzeugt das Schlüsselpaar lokal auf ihrem Computer, hinterlegt nur den öffentlichen Teil bei Albatros und testet den ersten Besuch auf dem Server. Nutze den Skill bei «SSH-Schlüssel», «ssh-keygen», «ich komme nicht auf meinen Server», «Permission denied (publickey)», «was ist ein public key», «neuer Computer», und als Schritt 4 von /albatros:start.
Gets Albatros started for someone with no technical background. Asks whether they want a website or a shop, guides them through signing up in the browser, choosing a plan (Free for learning, Light for a website from the Logbuch, Pilot with their own server), and setting up an SSH key and a site or server until the first page is live at name.aufwind.sh. Use this skill for /albatros:start, “start Albatros,” “I want a website,” “I have a business idea,” “how do I get started,” “I’m new here,” “do I need a server,” “how much does Albatros cost,” and when someone mentions Albatros but has no account or site yet.
Sets up privacy-friendly website analytics with PostHog alongside someone with no prior knowledge. First, create a small event plan (which actions to count, and exactly when), then install it, set up a goal event such as “inquiry sent,” filter out your own visits, and compare the results with actual inquiries. Use this skill for “analytics,” “count visitors,” “how many people visit my site,” “PostHog,” “Google Analytics,” “tracking,” “conversion,” “measure an event,” and “where do my visitors come from.”
Build, change, run and publish a Teable app (an App Builder app, Next.js on Teable data) from this machine with the `teable` CLI instead of the in-app chat. Use when the user wants to create a Teable app, work on an existing one locally, push a new version, publish it, roll back, or react to table-schema changes and named-query repairs. Covers project (base) resolution, `teable app init|clone|pull|push|publish`, the local dev loop and the platform rules the checkout must follow.
Use the installed pc CLI to query or modify PingCode projects, work items, product requirements, tickets, tests, knowledge bases, organizational data, and the Nexus application platform API. Use when the user wants to complete tasks in PingCode or configure PingCode CLI login.
Viết như người (human voice, Vietnamese humanizer): chống giọng văn AI cho tiếng Việt và tiếng Anh. Viết hoặc sửa văn bản cho giống người viết, bỏ các dấu hiệu văn AI. Dùng khi người dùng gọi /viet-nhu-nguoi, hoặc nhờ "viết như người", "viết cho giống người", "human voice", "humanize", "chống giọng văn AI", "bớt giọng AI", "bớt giọng ChatGPT", "viết tự nhiên hơn", "sửa văn AI", "soi giọng AI", hoặc khi soạn tin nhắn, email, bài viết, PR, commit, slide, lời thuyết trình mà người dùng sẽ gửi hoặc trình bày dưới tên mình. Có phần riêng cho tiếng Việt (văn dịch từ tiếng Anh, "việc/sự/một cách", giọng công văn, xưng hô) và cho slide (tiêu đề kết luận, ghi chú thuyết trình, quét thẳng file .pptx). Dựa trên Wikipedia "Signs of AI writing", skill humanizer (MIT) và các skill humanize tiếng Việt mã nguồn mở.
Use Virse CLI to understand canvas content and regions, read node configurations and relationships, place existing assets, and edit nodes, groups, connections, layouts, duplicates, restorations, and canvas discussions. For canvas structure and material selection; use virse-creation to generate or modify pixel content, and virse-assets to organize asset directories.
In Virse, choose an in-canvas creator Composer for the task, delegate canvas creation subtasks, manage independent nodes and sessions, pass node references, continue conversations, read history, observe turns, or request a stop. Suited to goals requiring multi-step creation and feedback within the canvas; for precise management or execution with a specified model or tool output, use the CLI directly.
Create or modify image, video, 3D, and text works in Virse; reuse existing materials and generation inputs; select a model or existing tool; and compare actual outputs and iterate. If the user explicitly specifies an existing smart tool, use the tool entry directly; use the canvas entry just to arrange node positions, the author entry to create reusable tools or aesthetics, and Composer for multi-step creation and feedback within the canvas.
Find, add, place, or run existing smart tools in Virse; read input forms for fixed outputs; observe and control runs; and manage personal lists, activation, favorites, and visibility settings the author can change. Use when the user asks to process materials with a specific tool or fixed version; use smart-tool-creator to write or modify tool package execution logic or build and publish.
Keyless multi-provider web search for agents. Twenty-one providers behind one command — eighteen classic ones plus an opt-in vendor tier (Keenable, Firecrawl, Exa) that returns page excerpts instead of teasers — with automatic fallback, query relaxation, cross-provider agreement ranking and domain diversity. Use for any "look this up", "what is the latest", "find me a library/tool/paper" request when you only need results, not a full report.
Plant mit einer Person ohne Vorwissen ihre Webseite, bevor gebaut wird: für wen, was Besucher tun sollen, welche Seiten es braucht, welche Inhalte schon da sind, welche Vorbilder gefallen. Ergebnis ist ein Grundriss in docs/plan.md. Nutze den Skill bei «ich will eine Webseite», «Webseite planen», «was brauche ich für eine Webseite», «welche Seiten», «Sitemap», «wo fange ich an», und immer bevor auf dem Weg Webseite die erste Seite gebaut wird.
Erklärt Snapshots mit dem Bild vom täglichen Foto der ganzen Wohnung, macht vor riskanten Änderungen ein eigenes Foto und holt den Server mit «Zurück auf gestern» wieder her. Enthält die Übung «absichtlich kaputt machen und reparieren». Nutze den Skill bei «Snapshot», «Backup», «ich habe etwas gelöscht», «Seite ist kaputt», «Server geht nicht mehr», «rückgängig machen», «zurück auf gestern», und bevor du auf dem Server etwas Grosses umbaust.
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.