prioritizing-my-inbox
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of inbox report titles and summaries which could contain malicious instructions.
- Ingestion points: Untrusted data enters the agent context via the
inbox-reports-listtool call inSKILL.md. - Boundary markers: The instructions do not specify any delimiters or warnings to the LLM to ignore embedded instructions within report content.
- Capability inventory: The skill uses
activity-summary-get,inbox-reports-list, andinbox-ranking-outcomes-list, but is restricted to read-only operations. - Sanitization: There is no mention of sanitizing or filtering the content of report titles or summaries before they are evaluated for scoring.
Audit Metadata