product-feed-optimizer
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill features an attack surface for indirect prompt injection as it ingests untrusted data from product feeds and landing pages. * Ingestion points: The skill reads user-supplied product feed exports (TSV/CSV/XML) and scrapes external landing pages as specified in the Data Sources section. * Boundary markers: The skill includes a dedicated security instruction to treat all external exports and scraped data as untrusted input. * Capability inventory: The skill writes processed data to local memory paths and utilizes a local script (registry-events.py) to log claim events. * Sanitization: Instructions explicitly command the agent to never follow instructions embedded within the processed data (SKILL.md, Instructions section).
Audit Metadata