A local macOS image analysis tool leveraging Apple's Vision framework to detect faces, text, poses, and objects without external API calls.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install vision-tagger
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install vision-tagger using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Vision Tagger?
Vision Tagger is a high-performance, macOS-native image analysis skill that utilizes the power of Apple's Vision framework. Designed for developers who prioritize privacy and speed, this tool performs all computations locally, meaning no images are ever sent to the cloud and no API keys are required. It serves as a robust addition to the ecosystem of Openclaw Skills by providing a bridge between system-level Swift APIs and accessible CLI or Python-based workflows.
The skill is capable of extracting highly granular data, from 76-point facial landmarks to 21 individual hand joints. Whether you are building an automated photo organizer, a document processing pipeline, or a posture-tracking application, Vision Tagger delivers structured, actionable JSON data alongside optional visual annotations for immediate verification.
Vision Tagger Use Cases
- Automated photo tagging by identifying over 1000 scene categories and objects.
- Posture and ergonomics monitoring via real-time body and face orientation tracking.
- Localized document scanning and data extraction using high-accuracy OCR.
- Privacy-centric security applications that detect human presence without cloud processing.
- Media accessibility enhancement through automated image description and saliency detection.
How Vision Tagger Works
- The user passes an image path to the compiled Swift utility.
- The utility initializes the Apple Vision framework to run concurrent detection requests.
- The system analyzes the image for faces, bodies, hands, text, barcodes, and scene labels.
- A structured JSON payload is generated, containing bounding boxes, joint coordinates, and confidence scores.
- Optionally, the data can be piped into a Python annotation script to create visual overlays for debugging or presentation.
Vision Tagger Setup
To get started with this entry in the Openclaw Skills collection, ensure you are running macOS 12+ and have the Xcode Command Line Tools installed.
# Install required build tools
xcode-select --install
# Install Python image processing library
pip3 install Pillow
# Compile the native Swift analysis binary
cd scripts/
swiftc -O -o image_tagger image_tagger.swift
Vision Tagger Data Schema & Taxonomy
Vision Tagger outputs a comprehensive JSON schema detailing every detected element in the image. Key sections include:
| Section | Data Points |
|---|---|
dimensions |
Integer width and height of the processed image. |
faces |
Bounding boxes, landmarks (76 points), and roll/yaw/pitch angles. |
bodies |
18 skeleton joints including neck, shoulders, elbows, and ankles. |
hands |
21 joints per hand with left/right chirality (handedness) identification. |
text |
OCR text strings, confidence scores, and precise bounding boxes. |
labels |
Scene classifications like 'office', 'outdoor', or 'furniture'. |
barcodes |
Decoded data from QR codes, UPC, EAN, and more. |
Vision Tagger Advanced Features
- Completely local execution ensures data privacy and zero latency from network overhead.
- Multi-modal detection allows for OCR, face tracking, and pose estimation in a single execution pass.
- High-precision skeletal analysis featuring 18 body joints and 21 hand joints for complex motion study.
- Saliency region detection to identify the most visually important parts of an image (attention-based and objectness-based).
- Flexible integration options allowing the tool to be called via shell, Python subprocess, or integrated into larger Openclaw Skills automation pipelines.
SKILL.md
Loading
Related Openclaw Skills
stdio-skill
A secure filesystem-backed bridge for exchanging files between users and AI agents via MCP.
Project Router
A terminal-first project bootstrapper and workspace context manager that synchronizes local task management with Trello.
Knowledge Graph
Maintain a deterministic, file-based knowledge graph by adding, superseding, and summarizing atomic facts for entities.
Finance Skill
A personal finance memory layer that enables AI agents to parse statements, store transaction history, and perform spending analysis.
MoltFundMe
An AI agent skill for discovering, evaluating, and advocating for decentralized crowdfunding campaigns while building a reputation through a karma system.
MoltFundMe Crowdfunding & Advocacy
MoltFundMe enables AI agents to discover, evaluate, and advocate for crowdfunding campaigns while earning reputation karma within a specialized ecosystem.
Featured*
