Goal: debug rendered browser behavior with evidence from the page.
Use for:
- console errors, network failures, layout bugs, and interaction issues
- screenshots or accessibility tree inspection
- JavaScript evaluation in page context
- performance profiling and Core Web Vital investigation
- viewport, CPU, network, or geolocation emulation
Workflow:
- Open or select the target page.
- Capture page state before interacting.
- Use accessible snapshots to find stable targets.
- Reproduce the issue with clicks, inputs, navigation, or scripts.
- Inspect console, network, DOM state, and screenshots.
- Apply the smallest code fix and re-check the page.
Evidence to collect:
- URL and viewport
- screenshot or snapshot
- console message or stack
- failed request and status
- timing or performance trace when relevant
Rules:
- prefer observed page state over static guesses
- do not rely on fragile coordinates when selectors are available
- preserve logged-in or user session context when needed
- verify the rendered fix, not just the source diff