multimodal-dataprep-user
Pass
Audited by Gen Agent Trust Hub on Aug 20, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches a setup script (
setup.sh) and Docker Compose configuration files from the author's public repository (github.com/open-edge-platform/edge-ai-libraries) to bootstrap the deployment environment. - [REMOTE_CODE_EXECUTION]: A remote setup script is downloaded and sourced locally to initialize environment variables. This is a standard deployment step for this tool, and the source is the author's official project repository.
- [COMMAND_EXECUTION]: Shell commands are used to manage Docker containers and verify service readiness. A health check uses
python3 -cto parse JSON response data from a local endpoint, which is a standard administrative practice for status validation. - [CREDENTIALS_UNSAFE]: Environment variable examples for MinIO storage use descriptive placeholders such as
<user>and<strong-password>. The documentation correctly advises against committing actual credentials to the environment. - [PROMPT_INJECTION]: The service ingests media files and directory paths. Documentation confirms that the tool includes validation to prevent directory traversal outside of the designated ingestion root, mitigating potential indirect injection or unauthorized file access via the ingest-dir endpoint.
Audit Metadata