deal-slip-scenario
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection Surface: The skill processes information from potentially untrusted external sources (such as emails, chats, and documents). While this is a standard pattern for its intended use case, it introduces a potential surface where external content could attempt to influence the agent's behavior.
- Ingestion points: Defined in
SKILL.mdas including email, chat, transcripts, enrichment data, and external documents. - Boundary markers: The skill includes explicit instructions to treat this content strictly as data rather than instructions and to report any instruction-like text detected within these sources.
- Capability inventory: As noted in
SKILL.md, the agent can perform actions like updating records, sending emails, and booking meetings via connected tools. - Sanitization: The instructions provide specific defensive measures, such as suppressing the rendering of external links and requiring manual user confirmation for any action that originates from untrusted content, ensuring consistent human-in-the-loop control.
Audit Metadata