| name | Web Interface Verification |
| description | This skill outlines the workflow for verifying the Redbox Portal web interface using the Agent's built-in browser capabilities and natural language test scripts. |
Skill: Web Interface Verification
Context
This skill outlines the workflow for verifying the Redbox Portal web interface using the Agent's built-in browser capabilities and natural language test scripts.
Workflow
1. Start the Application
Start the application in development mode with the mount profile. The dev:run script applies the mount profile implicitly (no extra flags required).
Important: Capture the CommandId returned by this command to monitor logs later. The CommandId is typically a UUID or numeric ID printed near startup; look for it before or around the "Sails lifted" readiness marker.
npm run dev:run
Example output (capture the value after CommandId:):
CommandId: 9f2b5c3a-1d4e-4a6a-9e5b-0a1234567890
...Sails lifted...
Extract it from the logs with a one-liner (or copy the CommandId: line directly):
grep -m 1 "CommandId:" .tmp/logs/dev-run.log
Wait for the application to be ready (look for "Sails lifted").
2. Launch Browser
Use the browser_subagent tool to launch a browser session.
3. Execute Natural Language Script
Follow the natural language instructions provided by the user.
Example Script:
- "Login to the system."
- "Expectation: You are navigated to
/researcher/home."
- "Navigate to 'My Data'."
- "Verify the 'Create Record' button is visible."
Agent Responsibilities:
- translate high-level actions (e.g., "Login") into interaction steps (type username, type password, click login).
- Verify URLs and UI elements match expectations.
4. Monitor Logs
While the browser session is active or after a step fails, check the logs of the running application using the command_status tool and the captured CommandId.
- Look for server-side errors (500s, exceptions) that correlate with UI actions.
5. Report Findings
After completing the script or encountering a failure, report the results to the user.
- Success: Confirm all steps passed.
- Failure: Detail which step failed, the observed vs. expected behavior, and any relevant server logs.