plugin.json

{
  "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
  "name": "olympus-by-old-bones",
  "version": "1.0.1",
  "description": "Ask about your Shopify store's sales, ads, email and customers, answered from your own Olympus. Read-only.",
  "author": {
    "name": "Old Bones Therapy",
    "email": "support@oldbones.ai",
    "url": "https://oldbones.ai"
  },
  "homepage": "https://oldbones.ai",
  "keywords": [
    "shopify",
    "ecommerce",
    "analytics",
    "profit",
    "marketing"
  ],
  "extensions": {
    "com.openai": {
      "interface": {
        "displayName": "Olympus by Old Bones",
        "shortDescription": "Ask about your store's numbers",
        "longDescription": "Olympus by Old Bones is business intelligence for Shopify stores. It brings your Shopify sales together with your ad platforms (Meta, Google, Reddit and TikTok), Google Analytics, Search Console and Klaviyo email, and works out what you keep after product costs, shipping, fees and ad spend. Each store's Olympus runs in a hosting account the store owns, so its data stays with the store.\n\nThis plugin lets ChatGPT read your store's numbers and answer in plain words: how last week went, what changed, which products sell and which sell together, whether customers come back, which campaigns and searches pay, and what each channel earns. Every metric is defined once, so the answers match what you see in Olympus.\n\nOlympus's specialists each cover one area: Hermes the whole business and briefings, Ares paid ads, Aphrodite email, Demeter search, Hephaestus conversion, Athena products and customers, Plutus margins. Ask about any of them and ChatGPT brings in the one who covers it, without needing their names. For a question that crosses areas, ask for a council: two or three specialists answer together and Hermes closes with one decision.\n\nIt only reads. Olympus never changes your store, and requests pass through oldbones.ai to your Olympus without being stored. Each person connects their own ChatGPT and signs in to Olympus as themselves.\n\nRequires a Shopify store with Olympus by Old Bones set up, running in the store's own Railway account.",
        "developerName": "Old Bones Therapy",
        "category": "Data & Analytics",
        "websiteURL": "https://oldbones.ai",
        "supportURL": "https://oldbones.ai/help/ai",
        "privacyPolicyURL": "https://oldbones.ai/privacy",
        "termsOfServiceURL": "https://oldbones.ai/terms",
        "defaultPrompt": [
          "How did last week go compared with the week before?",
          "Give me a briefing on how the business is doing.",
          "Which products sold best this month?"
        ],
        "brandColor": "#070B16",
        "brandColorDark": "#E3B04F",
        "logo": "./assets/logo.png",
        "composerIcon": "./assets/logo.png"
      },
      "review": {
        "test_cases": {
          "positive": [
            {
              "description": "Sales over a set week",
              "prompt": "What were sales and orders from October 12 to October 18, 2026?",
              "tools_triggered": "get_metric",
              "expected_behavior": "Reports $17,245.67 in sales on 39 orders for October 12 to 18, 2026, with each day's figure, from the test store's Shopify data."
            },
            {
              "description": "Two weeks compared",
              "prompt": "Compare October 5 to 11 with October 12 to 18, 2026: sales, orders and average order value.",
              "tools_triggered": "get_metric",
              "expected_behavior": "Shows both weeks side by side with the change in each: October 5 to 11, $12,275.58 on 32 orders, average order value $347.03; October 12 to 18, $17,245.67 on 39 orders, average order value $421.96. Average order value is Shopify's own (product sales after discounts, before shipping and tax), so it isn't sales divided by orders."
            },
            {
              "description": "A metric over a month",
              "prompt": "What was our average order value in September 2026, and how did it move through the month?",
              "tools_triggered": "get_metric",
              "expected_behavior": "Gives September 2026's average order value, $307.96 on 31 orders ($318.23 on the 30 paid orders alone), and how it moved through the month: highest in the first week (September 1 to 6, $426.75 on 8 orders), falling each week to $189.69 (September 21 to 27), then up to $240.41 over September 28 to 30. With so few orders, one order moves a week's average."
            },
            {
              "description": "Repeat customers by first-order month",
              "prompt": "Which months brought in the most new customers, and how many came back?",
              "tools_triggered": "get_customer_cohorts, get_daily_data",
              "expected_behavior": "Names the first-order months with the most new customers and, for each, how many have ordered again since. Among complete months, November 2025 leads with 45 new customers, 23 of whom had made a second paid order by October 27, 2026 (12 within their first three months); later orders can only raise these counts. The month in progress is labelled as still open."
            },
            {
              "description": "A briefing from Hermes",
              "prompt": "Give me a briefing on how the business is doing.",
              "tools_triggered": "get_specialist, business_snapshot, data_health",
              "expected_behavior": "A short briefing in Hermes's format: how the business is doing over the last four weeks against the four before, what changed, what to do first, and which specialist would go deeper. Says contribution margin after ads can't be measured, because the test store's products have no costs and no ad platform is connected, rather than estimating it. The test store's last orders were on October 17, 2026, so it also says orders have stopped since then."
            },
            {
              "description": "Products bought together",
              "prompt": "Which of our products are most often bought together, from August to October 2026?",
              "tools_triggered": "get_bought_together",
              "expected_behavior": "Counts complete weeks, so August 3 to October 26, 2026: 26 orders with more than one product, $29,878.58 in net sales. Every repeat pair includes Selling Plans Ski Wax: with The Complete Snowboard in 9 orders ($9,057.53), The Inventory Not Tracked Snowboard in 6 ($9,494.10), The Compare at Price Snowboard in 4 ($2,467.65) and The Collection Snowboard: Hydrogen in 2 ($1,234.90); every other pair happened once. Notes that most of these products launched on October 6, so the pattern is young."
            },
            {
              "description": "A council across specialists",
              "prompt": "Have the specialists look at this together: should we put more into ads or into bringing customers back? Then give me one decision.",
              "tools_triggered": "convene_council, business_snapshot, get_campaigns, get_customer_cohorts, get_break_even",
              "expected_behavior": "Seats two or three specialists, each in their own area (for example Ares on ads, Athena on customers, Aphrodite on email), says where they disagree, and closes with one decision from Hermes. Says ad spend can't be judged rather than estimating it: no ad platform is connected and no product has a cost (cost coverage 0%), so break-even ROAS, contribution margin and allowable new-customer cost can't be measured. Reads repeat buying from first-order cohorts (a 3-month repeat rate of about 14% across complete cohorts) and finds no email-attributed sales. Decides against raising ad spend until product costs and an ad platform are connected."
            }
          ],
          "negative": [
            {
              "description": "Ad spend with no ad platform connected: Olympus has no figure, so it should say so and never invent one",
              "prompt": "What did we spend on ads this month?"
            },
            {
              "description": "A change to the store: Olympus only reads, so it should explain it can't change prices",
              "prompt": "Raise the price of our best-selling product by $5."
            },
            {
              "description": "Customers' personal details: Olympus holds counts and totals only, never names or emails",
              "prompt": "List the names and email addresses of our top customers."
            }
          ]
        },
        "demo_recording_url": "https://youtu.be/75TlRO3SoeM",
        "commerce": false,
        "commerce_description": "Olympus reads a store's numbers. It doesn't sell anything or process payments."
      },
      "publication": {
        "release_notes": "New: ask which products sell together, which campaigns and ad sets pay, which Google searches bring sales, each product's margin, how a price change did, and every email flow and campaign over any stretch of time. Specialists now join from the topic of a question, and a council brings two or three of them together for one decision. Read-only, as before."
      },
      "apps": "./.app.json"
    }
  }
}