lemonade-router-builder
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The validation script (
scripts/validate.py) uses thesubprocessmodule to invoke the system'snodeinterpreter. This is used to verify that user-defined regular expressions are compatible with the target environment's requirements.\n- [PROMPT_INJECTION]: The skill processes untrusted natural-language input to generate structured routing policies, which creates a surface for indirect prompt injection.\n - Ingestion points: Natural-language descriptions of routing intent provided by the user (as described in
SKILL.md).\n - Boundary markers: The skill lacks explicit delimiters or instructions to ignore embedded instructions within the user-provided text.\n
- Capability inventory: The skill generates a local JSON file (
router.json) and executes a bundled validation script (scripts/validate.py) that performs sub-process calls.\n - Sanitization: The skill does not perform specific sanitization or filtering of the user's input before using it to generate configuration logic.
Audit Metadata