quick
Pass
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input via the $ARGUMENTS variable to determine code changes. Ingestion points: SKILL.md. Boundary markers: Absent. Capability inventory: File writing and execution of mix commands (mix compile, mix ecto.migrate, mix test). Sanitization: Absent.
- [DYNAMIC_EXECUTION]: The skill generates Elixir source code, database migrations, and schema updates at runtime based on user requests, which are then compiled and executed using the system's mix build tool.
Audit Metadata