device-guidance-v1-0-4
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection Surface: The skill processes data retrieved from an external tool, which introduces a potential surface where external content could influence the agent's behavior.
- Ingestion points: Data retrieved from the
Get Employee Assetstool is used to populate the device options table and define subsequent actions (SKILL.md). - Boundary markers: The skill lacks explicit delimiters or instructions to treat data from the asset tool as untrusted content, which is a common practice to mitigate instruction override risks.
- Capability inventory: The skill possesses the capability to send emails using the
Send device request emailtool based on the data processed from the asset list. - Sanitization: There is no evidence of sanitization or filtering of the content retrieved from the SharePoint source before it is rendered or used in tool calls.
Audit Metadata