resume-approvals-not-visible
Pass
Audited by Gen Agent Trust Hub on Apr 9, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by interpolating untrusted user data into diagnostic queries.
- Ingestion points: Placeholders for
{contract_id},{user_email}, and{request_id}inSKILL.mdused for database and log lookups. - Boundary markers: Absent; there are no delimiters around the interpolated data or instructions to the model to treat these values as literals.
- Capability inventory: The agent uses BigQuery SQL execution and GCP Cloud Logging tools to interact with production environments.
- Sanitization: No sanitization, escaping, or validation steps are defined for the input parameters before they are used in queries.
Audit Metadata