rerun-dataset-conversion
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion and processing of external multi-modal robotics datasets (MCAP, HDF5, LeRobot, etc.). While this represents a potential surface for indirect prompt injection, the instructions mitigate this risk by mandating thorough data surveys (Step 2), census checks for undecodable channels, and explicit human sign-offs at critical stages of the conversion mapping (Step 3) and blueprint design (Step 5).
- [EXTERNAL_DOWNLOADS]: The instructions guide the agent to assist the user in writing scripts for downloading datasets and uploading processed results to catalog storage (e.g., S3 or local Rerun servers). These operations are core to the skill's primary purpose and include safeguards such as asking for maximum download sizes and resolving local paths to absolute URIs for registration.
Audit Metadata