mcda-sales-comparison
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes local shell commands for environmental setup, path discovery, and workflow execution. Specific commands include 'find' for locating the plugin root directory, 'mkdir' for report folder creation, and 'python3' to run the calculation engine.
- [PROMPT_INJECTION]: The skill exhibits a potential surface for indirect prompt injection through the processing of external JSON data. 1. Ingestion points: Data is ingested from the file path specified in 'INPUT_JSON_PATH'. 2. Boundary markers: The calculator script parses input as structured JSON using the 'json' library. 3. Capability inventory: The agent context has access to shell execution, the 'Agent' tool for subagent dispatch, and file system read/write operations. 4. Sanitization: Input is validated for required fields and value ranges (e.g., price per square foot); however, string values from the JSON are interpolated into the final markdown report template.
- [EXTERNAL_DOWNLOADS]: The 'validation.py' script includes an optional dependency on 'jsonschema'. The implementation uses a try-except block for graceful degradation and is limited to standard, well-known libraries for data validation purposes.
Audit Metadata