smolevich/laviolab-atlas

{
  "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
  "name": "laviolab-atlas",
  "version": "0.3.0",
  "description": "Get a side project found: run existing SEO and AI-search audits and order them into one plan, plan content by query clusters, and read a weekly Search Console slice from exports. By LaVioLab.",
  "author": {
    "name": "Stanislav Shupilkin / LaVioLab",
    "url": "https://atlas.smolevich.com"
  },
  "homepage": "https://atlas.smolevich.com",
  "repository": "https://github.com/Smolevich/pet-project-atlas",
  "license": "MIT",
  "keywords": [
    "seo",
    "geo",
    "ai-search",
    "indexing",
    "search-console",
    "content-plan",
    "analytics"
  ],
  "extensions": {
    "com.openai": {
      "interface": {
        "displayName": "LaVioLab Atlas",
        "shortDescription": "Get a side project found",
        "longDescription": "LaVioLab Atlas helps the author of a small product or side project get it found in search and in AI answers. It connects ChatGPT to the Pet Project Atlas (atlas.smolevich.com), a playbook written from measured logs and search consoles, and adds three skills that work from your own reports and exports.\n\nTools. All four only read the published atlas and need no account.\n- search_playbook finds the sections of the atlas that answer a question and returns a short verbatim quote with a link to the exact section, in English or Russian.\n- get_page returns one atlas page in full, with the date it was measured and its sources.\n- list_sections lists the route, from checking demand to taking the first payment, with every page in it.\n- checklist returns the atlas checklist items for a topic such as robots.txt and AI crawlers, llms.txt, canonical URLs, Search Console or JSON-LD, each with the page that explains it, plus how to verify the result.\n\nSkills.\n- start takes the reports of two open-source audits, seo-audit and geo-audit, merges them into one list ordered by what blocks what (fetchable, allowed, indexed, findable, quotable, attributed), links each item to the atlas page that explains it, and names three things to do this week. Without both reports it says how to get them and stops.\n- content-plan turns a description of the product and an optional Search Console query export into query clusters, one page per cluster, and a writing queue ordered by intent and money. It never invents search volumes.\n- report reads Search Console Performance and Page indexing exports plus your own product export, and returns six numbers with the delta against the previous period, three conclusions and one action.\n\nLimitations: the tools read only the atlas; they do not crawl or audit your site and do not connect to Google or to any account. Every number in a report comes from a file you upload. It does not write pages for you. Made by Stanislav Shupilkin at LaVioLab; open source under the MIT licence.",
        "developerName": "Stanislav Shupilkin",
        "category": "Data & Analytics",
        "capabilities": [
          "Search the atlas playbook and quote it with links",
          "Checklists for robots.txt, llms.txt, schema and indexing",
          "Merge SEO and AI-search audit reports into one ordered plan",
          "Build a content plan from query clusters",
          "Weekly Search Console report from exports"
        ],
        "websiteURL": "https://atlas.smolevich.com/tools/skills/",
        "supportURL": "https://github.com/Smolevich/pet-project-atlas/issues",
        "privacyPolicyURL": "https://atlas.smolevich.com/privacy/",
        "termsOfServiceURL": "https://atlas.smolevich.com/terms/",
        "defaultPrompt": [
          "What does the atlas say about letting AI crawlers in through robots.txt?",
          "Here are my seo-audit and geo-audit reports for my site. What do I fix first?",
          "Make a content plan from my product description and this Search Console export."
        ],
        "brandColor": "#2f6f4f",
        "brandColorDark": "#a7d7bd",
        "composerIcon": "./assets/icon.png",
        "logo": "./assets/icon.png"
      },
      "review": {
        "test_cases": {
          "positive": [
            {
              "description": "A question about llms.txt is answered from the atlas with a quote and a link.",
              "prompt": "Is llms.txt worth publishing for my SaaS site? Answer from the LaVioLab Atlas.",
              "tools_triggered": "search_playbook",
              "expected_behavior": "Calls search_playbook with a query about llms.txt. The answer quotes the atlas section on whether llms.txt is worth writing and links to https://atlas.smolevich.com/geo/llms-txt-and-crawlers/ (with the section anchor)."
            },
            {
              "description": "A checklist for a technical topic comes back item by item, each with its explaining page.",
              "prompt": "Give me the LaVioLab Atlas checklist for robots.txt and AI crawlers.",
              "tools_triggered": "checklist",
              "expected_behavior": "Calls checklist with the topic robots.txt and AI crawlers. Lists the matching items, including serving User-agent: * with Allow: /, each with a link to the atlas page that explains it, and the verification steps from the AI crawlers page."
            },
            {
              "description": "The route of the atlas is listed in order.",
              "prompt": "Show me the sections of the Pet Project Atlas in order, with their pages.",
              "tools_triggered": "list_sections",
              "expected_behavior": "Calls list_sections. Shows the sections in route order (Start here, Demand, Indexing, AI search, Content, Distribution, Analytics, Money, Tools, Cases) with the page titles and links."
            },
            {
              "description": "One page is read in full and summarised with its link.",
              "prompt": "Open the atlas page indexing/why-google-does-not-see-you and summarise its steps.",
              "tools_triggered": "get_page",
              "expected_behavior": "Calls get_page with slug indexing/why-google-does-not-see-you. Summarises the steps of that page and links to https://atlas.smolevich.com/indexing/why-google-does-not-see-you/."
            },
            {
              "description": "A question in Russian is answered from the Russian pages of the atlas.",
              "prompt": "Почему Google не видит мой сайт? Ответь по атласу LaVioLab.",
              "tools_triggered": "search_playbook",
              "expected_behavior": "Calls search_playbook with lang ru. Answers in Russian, quoting the Russian indexing pages of the atlas and linking to URLs under https://atlas.smolevich.com/ru/indexing/."
            }
          ],
          "negative": [
            {
              "description": "Creative writing has nothing to do with the atlas.",
              "prompt": "Write a haiku about autumn leaves."
            },
            {
              "description": "Weather is not covered by the atlas.",
              "prompt": "What is the weather in Belgrade tomorrow?"
            },
            {
              "description": "A plain translation needs no atlas lookup.",
              "prompt": "Translate good morning into Spanish."
            }
          ]
        }
      },
      "publication": {
        "release_notes": "Adds a read-only MCP server over the Pet Project Atlas with four tools: search_playbook, get_page, list_sections and checklist. The start, content-plan and report skills are unchanged.",
        "countries": [],
        "translations": {
          "ru-RU": {
            "subtitle": "Чтобы проект находили",
            "description": "LaVioLab Atlas помогает автору небольшого продукта или пет-проекта добиться, чтобы его находили в поиске и в ответах AI. Он подключает ChatGPT к Pet Project Atlas (atlas.smolevich.com) — справочнику, написанному по замерам из логов и поисковых консолей, — и добавляет три скилла, которые работают с вашими отчётами и выгрузками.\n\nИнструменты. Все четыре только читают опубликованный атлас, аккаунт не нужен.\n- search_playbook находит разделы атласа, которые отвечают на вопрос, и возвращает короткую дословную цитату со ссылкой прямо на раздел, по-английски или по-русски.\n- get_page отдаёт страницу атласа целиком, с датой замера и источниками.\n- list_sections показывает маршрут — от проверки спроса до первой оплаты — со всеми страницами.\n- checklist возвращает пункты чек-листа атласа по теме: robots.txt и AI-краулеры, llms.txt, канонические адреса, Search Console, JSON-LD, — у каждого страница с объяснением, плюс как проверить результат.\n\nСкиллы.\n- start берёт отчёты двух открытых аудитов, seo-audit и geo-audit, сводит их в один список по порядку блокировок (доступен, разрешён, проиндексирован, находится, цитируется, атрибутирован), к каждому пункту даёт страницу атласа с объяснением и называет три дела на неделю. Без обоих отчётов говорит, как их получить, и останавливается.\n- content-plan превращает описание продукта и выгрузку запросов из Search Console в кластеры, одну страницу на кластер и очередь текстов по интенту и деньгам. Объёмы поиска не выдумывает.\n- report читает выгрузки Search Console (Эффективность и Индексирование страниц) и выгрузку из вашего продукта и возвращает шесть чисел с дельтой к прошлому периоду, три вывода и одно действие.\n\nОграничения: инструменты читают только атлас — ваш сайт они не обходят и не проверяют, к Google и к аккаунтам не подключаются. Каждое число в отчёте берётся из загруженного вами файла. Страницы за вас не пишет. Автор — Станислав Шупилкин, LaVioLab; открытый код под MIT."
          }
        }
      }
    }
  }
}