About this skill

Problem

In fixed-camera footage, it is easy to miss whether a person is present in a restricted area, how many people appear, or whether repeated visits and off-hours intrusion are happening. This skill turns local video, images, or network URLs into structured humanoid detection results, targeting parks, offices, and restricted zones that need continuous entry and exit monitoring.

How it works

The skill accepts a local video path via --input or a public URL via --url, and can limit the detection area with --detection-region using x1,y1,x2,y2. Core capabilities include person presence detection, headcount statistics, appearance frequency statistics, and intrusion alerts. Running scripts/human_detection_analysis.py produces detection results, risk notes, recommendations, and a report link; historical reports are queried through --list from the cloud service and rendered as a Markdown table.

Boundaries

This capability works best with fixed-camera monitoring video, supports mp4/avi/mov, and allows up to 10MB per file. The output is for security management reference only, and real enforcement should follow local procedures. Use only the bundled skill scripts, while identity handling is completed internally without requiring user-provided identifiers.

Screenshots

Use Cases

Park security duty staff review a CCTV clip and need to confirm whether anyone entered a restricted zone at night and generate an alert report.
Property safety engineers analyze an office corridor mp4 to count people in a defined area, track appearance frequency, and flag abnormal loitering.
Ops security analysts pass a public monitoring URL so the skill downloads it and outputs intrusion warnings with a report link.
Security operations query historical humanoid detection reports after a chosen date and verify Markdown table links.

Best For

Park security duty staff who need to confirm unauthorized presence in restricted zones from night clips and produce alert records.
Property safety engineers who need headcount, appearance frequency, and abnormal loitering statistics for office corridor video.
IT ops security analysts who need structured reports and report links from public monitoring URLs.
Security operations leads who need date-filtered lists of historical humanoid detection reports and verified links.