jev-fanout
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input via the 'state' field in the JSON request structure, which is then evaluated by an AI model. This creates a potential surface for indirect prompt injection. \n
- Ingestion points: The 'state' field within the JSON shape defined in SKILL.md. \n
- Boundary markers: There are no explicit delimiters or specific instructions for the agent to ignore potentially malicious content within the 'state' field. \n
- Capability inventory: The skill performs network requests to 'POST /v1/systemone' and executes the 'jev' CLI tool. \n
- Sanitization: No sanitization or validation logic is defined for the external data ingested through the 'state' field. \n- [COMMAND_EXECUTION]: The skill documentation references the execution of 'npx jev turn' for routing and input guards, which is a standard usage of a CLI tool provided by the service. \n- [EXTERNAL_DOWNLOADS]: The skill refers to patterns and documentation hosted on 'docs.typesafe.ai', which is the official documentation domain for the TypeSafe Jev service provider.
Audit Metadata