dataset-conversion

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The core workflow involves processing multi-modal robotics datasets (MCAP, HDF5, Parquet, etc.) from external sources, which constitutes an indirect ingestion surface.
  • Ingestion points: Dataset samples are downloaded or loaded during the survey and conversion steps (SKILL.md, Steps 1 & 2).
  • Boundary markers: The instructions do not specify the use of delimiters or specific warnings to the agent when interpreting dataset metadata.
  • Capability inventory: The skill utilizes file system writes for recording data and network access for downloading datasets and registering with a catalog server.
  • Sanitization: No specific sanitization protocols are mentioned, as the focus is on lossless data reflection.
  • [DYNAMIC_EXECUTION]: The guidelines involve generating and executing local Python and shell scripts to automate data downloading, conversion, and registration.
  • Evidence: Procedural steps in SKILL.md and template code in references/registering.md involve script creation and execution for data handling tasks.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 07:51 AM
Security Audit — agent-trust-hub — dataset-conversion