catsharp-galois
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOWCOMMAND_EXECUTION
Full Analysis
- COMMAND_EXECUTION (LOW): The skill defines 'just' commands (e.g., genesis-catsharp, galois-verify) that accept user-provided arguments. While intended for mathematical operations, these parameters are interpolated into command strings without documented sanitization.
- INDIRECT_PROMPT_INJECTION (LOW): The skill exhibits an injection surface by processing external inputs that influence agent reasoning and local execution. -- Ingestion points: Command-line arguments 'seed', 'here', 'elsewhere', and 'pitch-classes' defined in the 'Commands' section. -- Boundary markers: Not present; the skill lacks delimiters or instructions to ignore embedded commands within these inputs. -- Capability inventory: Ability to execute local 'just' recipes and perform complex mathematical transformations using Clojure and Julia logic. -- Sanitization: No input validation or escaping mechanisms are described for the processed arguments.
- PROMPT_INJECTION (SAFE): No attempts to override system prompts or bypass safety filters were detected.
- DATA_EXFILTRATION (SAFE): No evidence of sensitive file access or unauthorized network communication.
Audit Metadata