goose-product-photos
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill utilizes a predefined set of MCP tools (mcp__gooseworks__*) to interact with the GooseWorks backend. This architecture centralizes sensitive operations like AI image generation and data storage, reducing the attack surface within the agent skill itself.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data retrieved from brand contexts and product imports. Ingestion points: brand_get_context, list_brand_products, and get_product_import tools. Boundary markers: The instructions explicitly state to use context data for grounding and not to invent product facts. Capability inventory: The skill can initiate image generations via generate_product_photos and approve results. Sanitization: None specified in the instructions. This presents a surface for indirect injection if imported product data contains malicious text, though the backend image pipeline and human-in-the-loop approval process provide natural buffers.
Audit Metadata