Employee Emotion Fluctuation HR Report

Problem context

In high-pressure roles such as R&D and customer support, sustained mood decline can appear before visible resignation signals, but HR teams often lack a stable, non-invasive way to notice these shifts. This skill frames the task as producing anonymous support cues from observable office behavior, not as diagnosing who is emotionally unwell.

How it works

It is intended for fixed office-camera video, accepting local mp4/avi/mov files or URLs. Under an anonymization assumption, it analyzes expression and posture signals such as smile frequency, frowning, visual sighing, solo sitting, peer interaction, and desk posture. The core logic builds a 30-day individual baseline, then compares current behavior using relative deltas such as smile_delta, sigh_delta, and solo_sit_delta. Only when meaningful shifts persist for at least 3 business days does it output an anonymous ID and desk-coordinate level alert with a support suggestion, rather than a diagnosis or performance judgment.

Scope and cautions

It fits mid-to-large enterprises that want an internal HR wellness signal, especially in high-stress teams. Hard constraints matter: employee consent and required approvals must be in place, daily analyzable time should be at least 2 hours, and the output must not be used for performance reviews, promotion, or termination decisions. Communication should remain natural work support, ideally with EAP or mental-health resources, instead of telling employees that camera analysis flagged them as low-mood.