Install
$ agentstack add skill-commerce-agentic-agentic-commerce-skills-agentic-commerce-generate-ai-descriptions ✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
About
Purpose
Takes products flagged by [agentic-commerce-audit-descriptions](../../catalog-audit/agentic-commerce-audit-descriptions/SKILL.md) (or any product list) and rewrites their descriptions using Claude to inject the structured signals AI shopping agents need: use-case language, dimensions, materials, target persona — without losing the brand voice.
The skill is conservative by default: it does not overwrite existing descriptions in a single pass. It writes the rewrite to a draft field and shows the diff. The merchant explicitly approves the batch before publish.
Prerequisites
- Authenticated Shopify CLI session
- API scopes:
read_products,write_products ANTHROPIC_API_KEYenv var set (for the Claude call)
Parameters
| Parameter | Type | Required | Default | Description | |-----------|------|----------|---------|-------------| | store | string | yes | — | Store domain | | product_ids | array | no | — | Specific product IDs to rewrite. If omitted, rewrites all products with description-dimension score ⚠️ Mutations are gated behind dry_run: true AND publish: false by default. The skill is designed to fail safe: the worst case (running with no flags) just writes drafts to a metafield, not the live description. Explicit publish: true is required to overwrite live descriptions. > > Claude can hallucinate specs. The prompt explicitly says "do not invent specs that cannot be inferred from the title or tags". But it is not 100% reliable on niche products. Always review before publish: true on products where accuracy of dimensions/materials matters (regulated categories like cosmetics, supplements, electronics with safety certifications).
GraphQL Operations
# productUpdate:mutation — validated against api_version 2025-01
mutation ProductUpdateDescription($id: ID!, $descriptionHtml: String!) {
productUpdate(input: { id: $id, descriptionHtml: $descriptionHtml }) {
product {
id
descriptionHtml
}
userErrors {
field
message
}
}
}
# Alternative: write to draft metafield
mutation ProductDescriptionDraftWrite($id: ID!, $value: String!) {
productUpdate(
input: {
id: $id,
metafields: [
{ namespace: "agentic_commerce", key: "description_draft", type: "multi_line_text_field", value: $value }
]
}
) {
product { id }
userErrors { field message }
}
}
Reference
The "what to inject" logic implements the description-dimension rules from METHODOLOGY.md. Re-run the catalog audit after fixes to measure score uplift.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: commerce-agentic
- Source: commerce-agentic/agentic-commerce-skills
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet — be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.