shopify-store-skills
Pass
Audited by Gen Agent Trust Hub on Apr 13, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill's interactive setup guide (
site/src/pages/index.astro) contains a pre-defined prompt that instructs the AI agent to adopt a specific persona ("Act as a concierge") and explicitly suppresses standard command output ("DO NOT paste raw command output", "run these silently"). These are behavior-override and output-concealment patterns used for UX optimization. - [PROMPT_INJECTION]: Multiple skills ingest untrusted data from Shopify (e.g.,
skills/order-intelligence/shopify-admin-order-notes-and-attributes-report/SKILL.md). - Ingestion points: Data enters the agent context through the GraphQL
ordersandreturnsqueries fetchingnote,customAttributes, andreturnReasonNotefields. - Boundary markers: None implemented in the instructions.
- Capability inventory: The skill performs read-only operations and CSV generation (e.g., in
skills/order-intelligence/shopify-admin-order-notes-and-attributes-report/SKILL.md), but lacks sanitization. - Sanitization: External content is interpolated directly into the context without escaping or validation, presenting a surface for indirect prompt injection.
Audit Metadata