---

name: bd-test description: "Manage skills for visual understanding: register, list, invoke, and delete detection skills. Supports person detection, pedestrian counting, vehicle recognition, OCR, pose estimation, object tracking, and more." allowed-tools: Bash, Read, Write, Edit

Skill Registration and Usage

🎯 What This Tool Does

visual understanding platform. This tool enables you to:

• Register Skills - Register detection skills from platform locally
• Invoke Skills - Call registered skills with images or video frames
• Visualize Results - Draw bounding boxes, generate grid references, preview ROI/Tripwire
• Define Detection Areas - Use interactive workflows to define ROI (electronic fencing) or Tripwire (detection lines)

Supported Detection Types: Person detection, pedestrian counting, vehicle recognition, OCR, pose estimation, object tracking, etc.

🔧 Prerequisites

Get API Key

1. Activate trial package

📚 Usage Guide

Basic Workflow

# 1. List available skills (preset and registered)
node scripts/invoke.mjs --list

# 2. Invoke a skill
echo '{"input0":{"image":"photo.jpg"}}' | node scripts/invoke.mjs ep-xxxx-yyyy

# 3. Visualize or display results (optional)
node scripts/visualize.mjs photo.jpg '<result-json>'

Or register a skill first then use it:

# Then invoke
echo '{"input0":{"image":"photo.jpg"}}' | node scripts/invoke.mjs ep-xxxx-yyyy

Define Detection Areas

Need to define regions of interest (ROI) or detection lines (Tripwire)?

• ROI Workflow — Create electronic fencing, detect only in specified regions
• Tripwire Workflow — Draw detection lines, count crossing events

Both workflows include complete interactive steps and example dialogs.

View Complete Documentation

• Type Definitions — Detection, ROI, Tripwire data structures
• Visualization Guide — Display detection results, generate grid
• Video Frame Extraction — Extract frames from video for detection
• Grid Input System — Use grid coordinates to specify points
• Example Code — Complete workflow examples

💡 Common Tasks

Register a Skill

After obtaining an ep-id:

Invoke a Skill

echo '{"input0":{"image":"photo.jpg"}}' | node scripts/invoke.mjs <ep-id>

With ROI:

echo '{"input0":{"image":"photo.jpg","roi":{...}}}' | node scripts/invoke.mjs <ep-id>

With Tripwire:

echo '{"input0":{"image":"photo.jpg","tripwire":{...}}}' | node scripts/invoke.mjs <ep-id>

Preview ROI/Tripwire

Preview on image before invoking:

node scripts/visualize.mjs photo.jpg '[]' preview.png \
  --overlays '[{"kind":"ROI","name":"zone","points":[...]}]'

Generate Grid

Help users specify point locations using grid coordinates:

node scripts/show-grid.mjs photo.jpg grid.png

---