fictiv
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the Fictiv web application, such as page text and chat messages, which could potentially contain malicious instructions aimed at overriding the agent's behavior.\n
- Ingestion points: External data enters the agent's context from app.fictiv.com via browser tools, as mentioned in SKILL.md and references/ui-map.md.\n
- Boundary markers: Ground Rule 6 in SKILL.md explicitly directs the agent to treat page content as data and not as instructions.\n
- Capability inventory: The agent has the ability to place financial orders, share sensitive quote details via email, and modify account-level workspace settings.\n
- Sanitization: The skill relies on LLM directive adherence and a mandatory human-in-the-loop 'approval gate' for all spending and sharing actions rather than automated sanitization.
Audit Metadata