Shopify Admin Agentic Readiness Audit

Purpose

Runs a store-side "Agentic Commerce Readiness" scan — the same questions the public audit.40rty.ai audit asks, but answered from inside the Shopify Admin with full catalog data. It scores whether AI shopping agents (ChatGPT, Gemini, Perplexity, agentic checkout) can FIND, READ, and RECOMMEND the store's products, then prints a prioritized gap list where every gap names the sibling agentic skill that fixes it. Read-only — it changes nothing. Use it first (and on a schedule) to decide which remediation skills to run.

Prerequisites

• Authenticated Shopify CLI session (shopify auth login --store <domain>)
• Required API scopes: read_products, read_files, read_content (themes), read_online_store_pages

Parameters

All skills accept these universal parameters:

Skill-specific parameters:

Workflow Steps

1. OPERATION: shop — query Inputs: none Expected output: Shop name, primary domain, social sameAs links, and policy presence — feeds the identity + policy checks.

2. OPERATION: themes — query Inputs: roles: [MAIN], then theme.files(filenames: ["templates/robots.txt.liquid", "layout/theme.liquid", "assets/llms.txt", "templates/llms.txt.liquid"]) Expected output: Whether the published theme allows AI crawlers (robots), ships an Organization JSON-LD block, and serves an llms.txt — feeds discovery + identity checks.

3. OPERATION: metafieldDefinitions — query Inputs: ownerType: PRODUCT Expected output: Which structured attributes are defined (material, specs, features) — feeds the metafield-coverage check.

4. OPERATION: products — query (paginate to sample_size) Inputs: first: 250, fields: descriptionHtml, category, media, metafields, variants{ barcode, sku, price } Expected output: Per-product completeness — description length, image alt-text coverage, barcode/GTIN presence, category assigned, metafield population.

5. OPERATION: files — query Inputs: first: 50, query: "media_type:IMAGE" (sample) — corroborate alt-text coverage at the file level. Expected output: Alt-text fill rate across product media.

6. HTTP (no auth, optional): GET https://audit.40rty.ai/api/audit?domain=<primary domain> Inputs: the shop's primary storefront domain from step 1 Expected output: the outside-in view the Admin API can't give — robots/llms.txt as actually served, no-JS rendering, UCP/ACP endpoints, live Product JSON-LD. 200 = JSON with score and ranked fixes; a fix's fixSkill may be null, in which case use the step 7 pillar → skill mapping. 202 = scan running, retry the same URL in 30s (up to ~4 min) or skip. 429 = this IP's daily cap on fresh scans (reports under 24h old are still served), skip. Prefer its findings over the theme-file guesses in step 2 when they disagree — it reads the live storefront.

7. COMPUTE (no API): roll the findings into a 0–100 readiness score across five pillars — Discoverable (robots/llms.txt), Trusted (Organization schema, sameAs, policies), Readable (descriptions, alt text, JSON-LD fields), Structured (metafields, category, barcodes), Matchable (title/tag/metafield richness for intent) — and map each failing pillar to its fix skill.

GraphQL Operations

graphqlCopy code
# shop:query — validated against api_version 2025-01
query AgenticReadinessShop {
  shop {
    name
    myshopifyDomain
    primaryDomain { url }
    contactEmail
    shopPolicies { type body url }
  }
}

graphqlCopy code
# themes:query — validated against api_version 2025-01
query AgenticReadinessTheme {
  themes(first: 1, roles: [MAIN]) {
    nodes {
      id
      name
      files(filenames: [
        "templates/robots.txt.liquid",
        "layout/theme.liquid",
        "assets/llms.txt",
        "templates/llms.txt.liquid"
      ]) {
        nodes {
          filename
          body {
            ... on OnlineStoreThemeFileBodyText { content }
          }
        }
      }
    }
  }
}

graphqlCopy code
# metafieldDefinitions:query — validated against api_version 2025-01
query AgenticReadinessMetafieldDefs {
  metafieldDefinitions(first: 100, ownerType: PRODUCT) {
    edges { node { namespace key name type { name } } }
  }
}

graphqlCopy code
# products:query — validated against api_version 2025-01
query AgenticReadinessProducts($first: Int!, $after: String) {
  products(first: $first, after: $after) {
    edges {
      node {
        id
        title
        descriptionHtml
        category { id fullName }
        tags
        media(first: 10) {
          edges { node { ... on MediaImage { id image { altText url } } } }
        }
        metafields(first: 20) { edges { node { namespace key value } } }
        variants(first: 100) {
          edges { node { id sku barcode price } }
        }
      }
    }
    pageInfo { hasNextPage endCursor }
  }
}

graphqlCopy code
# files:query — validated against api_version 2025-01
query AgenticReadinessFiles($first: Int!, $after: String) {
  files(first: $first, after: $after, query: "media_type:IMAGE") {
    edges { node { ... on MediaImage { id alt } } }
    pageInfo { hasNextPage endCursor }
  }
}

Session Tracking

Claude MUST emit the following output at each stage. This is mandatory.

On start, emit:

Copy code
╔══════════════════════════════════════════════╗
║  SKILL: <skill name>                         ║
║  Store: <store domain>                       ║
║  Started: <YYYY-MM-DD HH:MM UTC>             ║
╚══════════════════════════════════════════════╝

After each step, emit:

Copy code
[N/TOTAL] <QUERY|MUTATION|HTTP>  <OperationName>
          → Params: <brief summary of key inputs>
          → Result: <count or outcome>

If dry_run: true, prefix every mutation step with [DRY RUN] and do not execute it.

On completion, emit:

For format: human (default):

Copy code
══════════════════════════════════════════════
OUTCOME SUMMARY
  <Metric label>:   <value>
  Errors:           0
  Output:           <filename or "none">
══════════════════════════════════════════════

For format: json, emit:

jsonCopy code
{
  "skill": "<skill-slug>",
  "store": "<domain>",
  "started_at": "<ISO8601>",
  "completed_at": "<ISO8601>",
  "dry_run": false,
  "steps": [
    {
      "step": 1,
      "operation": "<OperationName>",
      "type": "query | http",
      "params_summary": "<string>",
      "result_summary": "<string>",
      "skipped": false
    }
  ],
  "outcome": {
    "metric_key": 0,
    "errors": 0,
    "output_file": null
  }
}

Output Format

A readiness scorecard. human: an overall 0–100 score + per-pillar bars (Discoverable / Trusted / Readable / Structured / Matchable) + a prioritized gap table where each row is gap → impact → the agentic skill to run. json: { score, grade, pillars{...}, gaps:[{ pillar, audit_signal, finding, fix_skill }], sampled_products }. Every fix_skill value is a sibling skill name (e.g. shopify-admin-agentic-image-alt-text) so the operator can chain straight into remediation.

Error Handling

Best Practices

• Run this FIRST and re-run it after each remediation skill — it's the scoreboard that tells you what's left and what moved.
• Sample, don't crawl: 250 products is enough to estimate fill rates; only audit the full catalog when the sample shows borderline pillars.
• Treat category-unassigned and barcode-missing as the highest-leverage gaps — they unblock both AI retrieval (Matchable) and Product JSON-LD (Readable) at once.
• This skill is read-only; it never needs dry_run. The skills it routes you to DO mutate — run each of those with dry_run: true first.