om-auto-verify-pr-ui
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill defines a clear 'Untrusted content boundary' which instructs the agent to treat all data from pull requests, diffs, and comments as data to be analyzed rather than instructions to be followed. It specifically highlights patterns like 'ignore previous instructions' as indicators of malicious activity to be reported.
- [COMMAND_EXECUTION]: The skill uses shell commands to manage test environments and run Playwright. It enforces security best practices by requiring regex validation for branch names and numeric validation for PR numbers, ensuring all interpolated values are correctly quoted to prevent command injection.
- [DATA_EXFILTRATION]: There are explicit instructions to prevent the leakage of credentials. The skill mandates the use of demo credentials, prohibits the inclusion of secrets or tokens in screenshots, and requires the redaction of any sensitive information found during the UI verification process.
- [REMOTE_CODE_EXECUTION]: To mitigate risks from malicious code embedded in PRs, the skill requires the agent to generate its own Playwright test specifications based on observed UI changes. It expressly forbids the adaptation or execution of test code found within the repository or external comments.
Audit Metadata