Agentic Llms Txt

Purpose

llms.txt is the emerging convention for telling AI assistants, in plain text, what a site is and where the important pages live — a curated front door for models. This skill assembles an llms.txt from the store's own data (brand one-liner, top collections, flagship products with URLs, shipping/returns policy links, contact) and publishes it through a theme template so it's served at a stable path. Fixes llms-txt / well-known-llms-txt and the related agents-md discovery signals.

Prerequisites

Parameters

All skills accept these universal parameters:

Skill-specific parameters:

Safety

> ⚠️ Step 4 (themeFilesUpsert) writes a new template to the LIVE theme. The content is additive (a new file/route) and low-risk, but it is published immediately. Defaults dry_run: true so you review the generated llms.txt first. Duplicating the theme before writing is recommended.

Workflow Steps

Inputs: none

Expected output: Shop name, domain, contact email, policy URLs — the header + contact + policy section of llms.txt.

Inputs: first: flagship_count, optional query: "collection_id:''" (when flagship_collection_id is set — the curated set is your flagship list), sortKey: UPDATED_AT, reverse: true otherwise; fields title, onlineStoreUrl, description.

Expected output: Flagship product list with URLs for the "Key products" section. (Top-level products has no best-selling sort — use the curated collection for true bestsellers, else most-recently-updated as the proxy.)

Inputs: themeId, write templates/page.llms.liquid (a page template that outputs the text as text/plain) plus, if used, an assets/llms.txt copy; skipped when dry_run.

Expected output: Upserted theme file(s); collect userErrors. (Note: surface a one-line instruction to create a Page using the llms template, or to add a redirect from /llms.txt.)

GraphQL Operations

# shop:query — validated against api_version 2025-01
query LlmsTxtShop {
  shop {
    name
    primaryDomain { url }
    contactEmail
    shopPolicies { type url }
  }
}

# products:query — validated against api_version 2025-01
# Top-level `products` has no BEST_SELLING sort key. Pass a `collection_id:'…'`
# query filter to use a curated flagship set, otherwise fall back to the most
# recently updated products.
query LlmsTxtFlagship($first: Int!, $query: String) {
  products(first: $first, query: $query, sortKey: UPDATED_AT, reverse: true) {
    edges {
      node { id title onlineStoreUrl description }
    }
  }
}

# themeFilesUpsert:mutation — validated against api_version 2025-01
mutation LlmsTxtUpsert($themeId: ID!, $files: [OnlineStoreThemeFilesUpsertFileInput!]!) {
  themeFilesUpsert(themeId: $themeId, files: $files) {
    upsertedThemeFiles { filename }
    userErrors { filename code message }
  }
}

Session Tracking

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

On start, emit:

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

After each step, emit:

[N/TOTAL] <QUERY|MUTATION>  <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):

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

For format: json, emit:

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

Output Format

human: the full generated llms.txt + the post-publish step (create a Page on the llms template, or add a /llms.txt → page redirect). json: { theme_id, llms_txt, wrote: true|false, follow_up, errors }.

Error Handling

Best Practices