claudex-route
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute external CLI binaries (e.g., for GPT or Claude providers) to perform task handoffs. It explicitly mandates safety measures such as passing prompt text through stdin or files instead of shell interpolation, using temporary directories for process artifacts, and applying timeouts to child processes.
- [INDIRECT_PROMPT_INJECTION]: The skill processes data returned from external model CLI calls, which introduces a potential surface for indirect prompt injection.
- Ingestion points: The skill reads the 'completed result and exit status' from child CLI sessions in SKILL.md.
- Boundary markers: The instructions do not specify explicit delimiters for parsing results returned by sub-models.
- Capability inventory: The skill utilizes subprocess calls to external CLIs and has permissions for scoped file reads and writes as described in SKILL.md.
- Sanitization: The skill requires the agent to 'assess findings against evidence, inspect any edits, and run appropriate checks' before reporting outcomes to the user.
Audit Metadata