product-teardown
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill researches external products via URLs and reads project-specific documents, creating a surface for indirect prompt injection where malicious content on a researched site could influence agent behavior.
- Ingestion points: Processes product names or URLs for research and reads files at "{Product}/Product/teardowns/", "context/documents/index.md", and "context/learnings/".
- Boundary markers: The skill relies on standard Markdown output structures (headers and tables) to delimit sections but lacks explicit "ignore embedded instructions" warnings for processed data.
- Capability inventory: The skill performs file reads (FIND/UPDATE modes), file writes (CREATE/UPDATE modes to teardown directories), and network research (implied via instructions to research pricing, blogs, and changelogs).
- Sanitization: No specific sanitization or validation of external web content is described in the teardown methodology.
Audit Metadata