amc-run-video-calibration
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests configuration data from local JSON files (e.g., settings.json) to configure the calibration engine. This represents a potential injection surface for untrusted data.\n
- Ingestion points: The Python execution script reads configuration and alignment files from the user-specified videos directory.\n
- Boundary markers: No specific delimiters or warnings are used to wrap the ingested content in the file scanning process.\n
- Capability inventory: The skill has capabilities to read and write within the project workspace and perform network requests to the calibration microservice.\n
- Sanitization: The script sanitizes the 'detector_type' parameter by validating it against a whitelist of expected values ('resnet', 'transformer') before including it in API calls.\n- [COMMAND_EXECUTION]: The skill includes instructions to locate and run a bundled Python script for the calibration workflow. This execution is performed using standard system interpreters and is limited to the script file included with the skill package within its specific installation paths.
Audit Metadata