Sleep Quality Analysis

Problem It Addresses

Sleep-quality assessment often depends on wearables or clinical polysomnography. In home or lightweight medical settings, a more practical input may already be a monitoring camera video. This skill focuses on converting such sleep-monitoring video into readable findings: whether the subject entered deep sleep and REM, how often nighttime awakenings occurred, whether movement was excessive, and whether respiratory pause risk should be flagged.

How It Works

Core capabilities include:
- Sleep-stage recognition: distinguishes sleep onset, light sleep (N1/N2), deep sleep (N3), and REM
- Movement and awakening metrics: reports nighttime movement and awakening counts in a structured form
- Respiratory pause screening: highlights high-risk abnormalities such as OSAHS-related concerns
- Historical report lookup: reads past reports from the cloud API via --list and returns them as a Markdown table

A typical workflow is:
1. Provide a local video path or remote URL; supported formats include mp4, avi, and mov
2. Run the bundled skill script from the skill root
3. Use --input, --url, or --list
4. Control output depth with --detail and save results with --output

The skill enforces that historical reports come only from the cloud API, not from local memory, manual summaries, or long-term context. User identity is handled internally and should not be asked for or exposed.

Boundaries and Notes

This skill is intended as a sleep-quality reference for home or institutional use, but it does not replace professional sleep-monitoring devices or physician diagnosis. A video of at least 30 minutes is recommended so the sample can reflect a more complete sleep cycle. Remote URLs are downloaded by the API service, so the caller does not need to save them locally first. Outputs are for health reference, not clinical confirmation.