account-dossier
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from company websites, LinkedIn profiles, and job boards which could potentially contain hidden instructions designed to manipulate the agent's behavior.
- Ingestion points: Web fetcher and Apify scraper tools used in Phase 1 and Phase 3 to read external site content.
- Boundary markers: Uses a highly structured output template (Sections 1-11) which helps separate external data from instructional context, though no explicit 'ignore instructions' delimiters are used during processing.
- Capability inventory: Performs web searches, data scraping, and calls auxiliary skills for lead scoring and text humanization.
- Sanitization: There is no evidence of sanitization or escaping of the fetched external content before it is integrated into the final report or analyzed for fit.
Audit Metadata