convex-monitor
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from potentially untrusted sources like local error logs and deployment subscriptions.
- Ingestion points: Data enters through the
wait_for_eventtool as described inSKILL.md. - Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the event data are specified.
- Capability inventory: The instructions direct the agent to "fix" errors or "build" features based on event input, which implies the presence of file-writing and execution capabilities in the agent's environment.
- Sanitization: There is no mention of sanitizing or validating the contents of the events before the agent processes them.
Audit Metadata