About this skill

Problem

Child climbing on windows, balconies, or railings is a high-consequence home safety risk that is hard to monitor continuously from fixed camera feeds, especially in high-rise homes, kindergartens, and child-activity centers. This skill turns the footage into a dangerous-posture detection task, focusing on actions such as climbing, leaning out, stepping over barriers, or gripping window-sill edges.

How it works

Input: accepts a local file path or network URL; recommended formats are mp4, avi, and mov, with a maximum size of 10 MB and at least 15 FPS.
Execution: run the provided script from the skill root with --input or --url, and use --detail json to obtain structured results.
Detection: combines child detection and pose estimation to assess confidence for climbing, leaning out, barrier crossing, and unstable postures, then emits warning, critical, or emergency alerts.
Evidence: can generate an on-site snapshot and return a structured report with recommendations and a report link.
History lookup: uses --list to query cloud API results instead of relying on local memory or manually compiled records.

Boundaries

The output is an auxiliary early-warning signal, not a substitute for adult supervision. If an emergency alert triggers, respond immediately. Obtain guardian consent before use and store minor-related footage securely.

Screenshots

Use Cases

Use window camera video to detect child climbing and alert
Check railing video for child crossing or gripping
Integrate climbing alerts into smart-home security system
Validate footage for leaning, crossing, and unstable poses

Best For

Parents of young children who want window footage turned into reviewable climbing alerts.
Kindergarten safety leads who need railing footage checked for structured risk notices.
Smart-home integrators who need climbing detection wired into security systems.
Video safety testers who need stable detection of leaning, crossing, and unstable poses.