i4h-catheter-navigation-e2e
Pass
Audited by Gen Agent Trust Hub on Jul 22, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill clones the workflow repository from 'https://github.com/isaac-for-healthcare/i4h-workflows' if the local directory is missing. This is a legitimate download of the project's source code from the vendor's public GitHub repository.
- [COMMAND_EXECUTION]: Several bash commands are used to initialize the environment, including 'git clone', 'mkdir', 'ln', and 'export'. These are standard operations for setting up a development workflow and logging system.
- [PROMPT_INJECTION]: The skill represents a multi-step orchestration chain that processes external data (CT subjects via the 'SUBJ' variable) and passes it through multiple stages involving shell execution. There are no explicit boundary markers or sanitization steps in the skill's wrapper logic to prevent malicious instructions embedded in the data from influencing the agent's behavior.
- Ingestion points: 'SUBJ' environment variable in 'SKILL.md'
- Boundary markers: Absent
- Capability inventory: Subprocess calls (bash), file-write (logs), network ops (git) in 'SKILL.md'
- Sanitization: Absent in the skill instructions
Audit Metadata