code-testing-agent
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFE
Full Analysis
- Safe Reporting Practices: The skill includes specific instructions to ensure test names and report artifacts do not contain unsafe characters or control sequences. This serves as a security consideration to prevent potential metadata injection issues in CI/CD platforms or test runners that visualize these results.
- Network and Service Isolation: It mandates that generated tests must not call external URLs, bind to network ports, or require service discovery. By requiring all external dependencies to be mocked, the skill ensures that test execution remains local and isolated from the network.
- Non-Destructive Workspace Management: The agent is explicitly instructed to treat the current workspace as authoritative and avoid using destructive commands such as
git clean,git restore, orgit reset. It also manages its internal state in non-stageable directories to prevent working artifacts from being accidentally committed to version control. - Policy and Constraint Adherence: The execution contract prevents the agent from attempting to bypass system-level denied operations or using aliases to circumvent security restrictions. This ensures the agent operates strictly within the permissions granted by the host environment.
- Convention Discovery: The skill analyzes the existing codebase to understand naming and framework patterns before generation, ensuring that new tests integrate naturally with the existing project structure and security practices.
Audit Metadata