Video report

Use FFmpeg for every recording. Capture the smallest useful surface while the existing interaction tool drives the application. Deliver observed behavior, not a substitute for tests or separate performance measurements.

1. Define the proof

Identify the application and version or working state, initial conditions, reproduction steps, and expected visible result. Exercise the real interaction path: click the affected button rather than calling its handler directly. Use test data and obtain required approval before consequential actions.

Record the corrected state by default. Include the faulty state only if already recorded or safely accessible without disturbing the working tree. Label the comparison before/after or after-only; distinguish fixtures from product runs.

Done when the scenario has concrete actions and an observable success criterion.