amc-run-video-calibration
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a bundled Python script (
scripts/run_video_calibration.py) to manage the calibration workflow. It uses a bash block inSKILL.mdto locate the script across various possible installation paths and invokes it using the localpython3interpreter. - [DATA_EXFILTRATION]: The skill transmits local data, including MP4 video files, JSON configuration files, and images, to a user-specified REST API endpoint (
BASE_URL). This behavior is the intended and documented primary function of the skill for remote processing of datasets. - [INDIRECT_PROMPT_INJECTION]: The skill scans for and processes local configuration files which could influence the agent's workflow.
- Ingestion points: The
VIDEO_DIRand its parent directories are scanned forsettings.json,config.json, andalignment_data.json. - Boundary markers: No explicit prompt boundaries or "ignore" instructions are used when interpolating file-derived data into the workflow logic.
- Capability inventory: The skill has file system read access and network transmission capabilities via the
requestslibrary. - Sanitization: Content from JSON files is parsed using standard libraries, and specific fields (like
detector_type) are used to parameterize the API calls.
Audit Metadata