i4h-workflow-dataset-teleop
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill clones the Isaac for Healthcare workflows repository from
https://github.com/isaac-for-healthcare/i4h-workflows. This source is an official vendor resource associated with the skill author (nvidia). - [COMMAND_EXECUTION]: The skill executes multiple shell commands, including
git clone,./run.sh,./stop.sh, anduv run. These commands are used to manage the simulation environment and demonstration recording workflow. - [REMOTE_CODE_EXECUTION]: The skill executes a local script (
./run.sh) that is downloaded as part of the repository cloning process. This represents the execution of code fetched from a remote source at runtime, though the source is vendor-controlled. - [INDIRECT_PROMPT_INJECTION]: The skill processes HDF5 demonstration datasets (
demos.hdf5) using inspection and action analysis tools. - Ingestion points: Demonstration data recorded in HDF5 format is read by the
i4h-datasettool inSKILL.md. - Boundary markers: No specific boundary markers or "ignore instructions" warnings are documented for the data inspection process to mitigate potential instructions embedded in metadata.
- Capability inventory: The skill has capabilities for repository cloning (
git clone), executing simulation scripts (./run.sh), managing background processes (./stop.sh), and running project-specific tools viauv run. - Sanitization: There is no mention of sanitization or validation of the HDF5 content before it is processed by the inspection tools.
Audit Metadata