convex-insights
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of application logs and insights, which creates a theoretical surface for indirect prompt injection attacks. \n
- Ingestion points: External data enters the agent's context through the output of the
logsandinsightstools. \n - Boundary markers: The instructions do not mandate the use of delimiters or specific markers to isolate processed log content from the agent's instructions. \n
- Capability inventory: The skill is restricted to read-only observability tools (
logs,insights,functionSpec,status). It does not include capabilities for file system writes, arbitrary shell command execution, or external network requests beyond the provided tools. \n - Sanitization: The instructions focus on filtering and aggregating data client-side but do not specify procedures for sanitizing or escaping the content of the log lines. \n- [COMMAND_EXECUTION]: The skill uses official Convex MCP tools for application monitoring. These are scoped, vendor-provided tools designed for retrieving log and status information and do not facilitate arbitrary command execution.
Audit Metadata