openvmm-ci-investigation
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- Command Execution: The skill executes the local script
repo_support/investigate_ci.pyand uses the GitHub CLI (gh) to automate diagnostic tasks. This is a functional requirement for CI investigation but involves granting the agent shell execution capabilities within the repository. - Indirect Prompt Injection Surface: The agent parses external data from CI artifacts (JUnit XML, petri logs), which is a security consideration as these sources could contain unvalidated content.
- Ingestion points: Content is ingested from JUnit XML and JSONL files stored in the
/tmpdirectory after being downloaded from the repository's CI runs. - Boundary markers: The skill does not define specific boundaries or markers to isolate the analyzed log text from the agent's internal instruction set.
- Capability inventory: The agent utilizes the
ghtool and Python execution, providing it with network access and local script execution power. - Sanitization: Log messages are extracted and presented to the agent without specific sanitization against instructions that might be embedded in the test failure messages.
- External Resource Access: The skill downloads test artifacts from the project's official GitHub repository, which is the expected source for this utility.
Audit Metadata