paidf-auto-labeling

Pass

Audited by Gen Agent Trust Hub on Sep 1, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements a robust 'confirmation first' policy, requiring the agent to verify all critical inputs such as data paths, output directories, and model endpoints with the user before any operations are initiated.
  • [SAFE]: Subprocess execution is strictly controlled through the workflow-runner service; instructions mandate that the agent must present the exact command to the user and obtain explicit approval before execution.
  • [SAFE]: The skill follows industry-standard secret management practices by using environment variable placeholders (e.g., NVIDIA_API_KEY) and explicitly forbidding the hardcoding of tokens, API keys, or absolute home paths in repository files.
  • [SAFE]: Network activity is limited to health checks and status queries of user-provided VLM/LLM endpoints, with clear warnings against including sensitive credentials in logged URLs or command-line arguments.
  • [SAFE]: The architecture uses a stable 'scene directory' contract (DAFT layout) which ensures data processing occurs within predictable, user-defined boundaries and utilizes read-only mounts for source assets.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 1, 2026, 04:17 AM
Security Audit — agent-trust-hub — paidf-auto-labeling