Overview
This agent acts as an expert debugger specializing in diagnosing failures within iOS UI automation test skills generated by tools like mirroir-mcp. It ingests a structured JSON payload detailing the failed skill, the steps that failed, and preliminary OCR analysis.
The primary goal is to move beyond simple error reporting and provide deep, actionable insights into why the automation failed, considering complex mobile behaviors like layout shifts, timing issues, and navigation errors.
Capabilities
- Root Cause Analysis: Identifies underlying systemic problems such as incorrect screen context, iOS version incompatibility, or dynamic content interference.
- Concrete Fix Suggestion: Generates specific, actionable remediation steps, including precise coordinate adjustments (e.g.,
tapX, should_be), necessary timing modifications (delays), or required structural edits to the YAML skill file.
- Confidence Scoring: Provides a confidence level (
high, medium, low) alongside the diagnosis for immediate triage by developers.
- Structured Output: Guarantees output in a strict JSON format, making it directly consumable by CI/CD pipelines or reporting tools.
Example Use Cases
- Layout Shift Debugging: A test fails because an element's label is present but its coordinates have shifted due to a new OS feature (e.g., adding a notification banner). The agent will pinpoint the coordinate change and suggest updating the target coordinates.
- Timing Issue Resolution: A
tap action fails because the subsequent animation hasn't settled, causing the next element to be momentarily unavailable. The agent will recommend inserting an explicit wait_for step before the failing action.
- Navigation Error Correction: The test incorrectly assumes it is on the main settings screen when a previous navigation step failed silently. The agent diagnoses this context mismatch and suggests adding a necessary intermediate navigation or assertion to confirm the correct view state.