price-hunt
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by ingesting and processing content from numerous untrusted external websites.
- Ingestion points: The skill fetches and reads data from retail websites, price comparators (e.g., Geizhals, Idealo, Knibble, Tweakers, Keepa), second-hand marketplaces (e.g., Marktplaats, Kleinanzeigen, leboncoin), and customs authority websites as described in Steps 3, 4, and 5.
- Boundary markers: There are no explicit delimiters or instructions to the agent to isolate the fetched external content or to ignore any malicious instructions that might be embedded within the scraped web pages.
- Capability inventory: The agent can search a sensitive 'vault' using the
hatchdoortool (Step 1), dispatch subagents with specific briefs (Step 4), and offer to record new findings into a 'decision note' in the vault (Step 6). - Sanitization: No sanitization, validation, or structured parsing of the scraped HTML/text is performed before the model processes the data to generate its report and recommendations.
Audit Metadata