omh-verification-gate
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes shell commands such as 'git worktree' and 'gh api' to inspect repository state and fetch CI status metadata. It also executes local build, lint, and test suites as defined by the repository being analyzed to gather verification evidence.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted data including pull request diffs, commit messages, and automated tool outputs to produce a verification verdict. This creates a surface for potential indirect prompt injection. 1. Ingestion points: Pull request diffs, commit messages, and CI status reports (SKILL.md). 2. Boundary markers: The skill defines a 'not-evidence boundary' to separate descriptive claims from observed execution results. 3. Capability inventory: Shell execution of Git, GitHub CLI, and repository-specific build/test tools. 4. Sanitization: No explicit sanitization or escaping of interpolated external content is documented within the skill logic.
Audit Metadata