multimodal-dataprep-user

Pass

Audited by Gen Agent Trust Hub on Aug 22, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches setup scripts and Docker Compose configurations from the official vendor repository on GitHub (github.com/open-edge-platform). These resources are used for initial stack deployment and are documented as standard procedure.
  • [REMOTE_CODE_EXECUTION]: A multi-line command pipes a local health check response to a Python one-liner for JSON parsing. The Python code is a static string provided within the skill for status verification and does not execute remote or untrusted code.
  • [CREDENTIALS_UNSAFE]: The instructions use clear placeholders such as <user> and <strong-password> for environment variable configuration, which is a recommended practice to avoid hardcoding secrets.
  • [INDIRECT_PROMPT_INJECTION]: The skill provides an interface for ingesting external media files and text summaries. It explicitly mentions path traversal protections (MM_DATAPREP_INGEST_DATA_ROOT) to ensure that directory ingestion is restricted to authorized paths, reducing the attack surface for malicious data input.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 22, 2026, 12:05 AM
Security Audit — agent-trust-hub — multimodal-dataprep-user