agent-spec-generator
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection (LOW): The tools catalog defines an environment that ingests untrusted data from multiple external sources, creating a surface for indirect prompt injection attacks.\n
- Ingestion points: Glean Search, Glean Document Reader, Gmail, Google Docs, and Confluence (references/tools-catalog.md).\n
- Boundary markers: Absent in the provided tool configuration templates.\n
- Capability inventory: Includes automated actions like sending emails, posting to Slack, and creating Jira/Salesforce records (references/tools-catalog.md).\n
- Sanitization: No mention of sanitization or validation for external content inputs.\n- Privilege Escalation (LOW): The catalog documents the 'skipUserInteraction' flag which allows for the execution of sensitive operations without manual approval, increasing the risk associated with successful prompt injections.
Audit Metadata