i4h-workflow-e2e
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads project code and driver scripts from the official NVIDIA-managed
isaac-for-healthcareGitHub organization. - [REMOTE_CODE_EXECUTION]: The skill performs a
git cloneof a remote repository and subsequently executes shell scripts (run.sh) retrieved from that source. As the source is a trusted vendor organization, this operation is part of the intended functionality. - [COMMAND_EXECUTION]: The skill utilizes shell commands to manage the repository environment, perform dry-run validations, and execute the end-to-end pipeline in the foreground.
- [INDIRECT_PROMPT_INJECTION]: The skill accepts a
<workflow>parameter from the user which is interpolated into shell command strings, creating a potential injection surface if malicious input is provided. - Ingestion points: The
<workflow>placeholder in the execution instructions withinSKILL.md. - Boundary markers: No explicit delimiters or protective markers are used to isolate the parameter in the shell command templates.
- Capability inventory: Access to shell execution (
./scripts/e2e/run.sh), file system operations (git clone), and cleanup tools (stop.sh). - Sanitization: The provided instructions do not include explicit sanitization or validation logic for the workflow identifier prior to shell interpolation.
Audit Metadata