Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
A collection of three sub-skills for detecting AI-generated images: local EXIF and artifact analysis, ML-based classification using pre-trained HuggingFace models, and cloud-based detection via the Hive Moderation API. Each sub-skill runs independently or in combination using a majority-voting strategy. Supports detection of images from Stable Diffusion, Midjourney, and DALL-E.
Combining local inference, pre-trained ML models, and a cloud API in one package lets you choose between speed, accuracy, and privacy without switching tools.
When to use it
- Checking whether a social media profile photo is AI-generated
- Verifying image authenticity before publishing a news article
- Filtering AI-generated or NSFW images from a user upload pipeline
- Extracting EXIF metadata from a suspicious image to spot inconsistencies
- Running fully local detection without sending images to any external service
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Is this image AI-generated?
- 1Load the image and run image-analyzer to extract EXIF metadata and flag visual artifacts such as distorted hands or blurred backgrounds
- 2Pass the image to hf-image-detector for ML-based classification using the umm-maybe/AI-image-detector model
- 3Optionally send to hive-api for a cloud-based second opinion with NSFW filtering included
- 4Aggregate results from active sub-skills using the configured voting strategy
- 5Return a detection verdict with per-model confidence scores and any flagged anomalies
Detection result indicating whether the image is AI-generated, with confidence scores from each sub-skill and a summary of detected artifacts or metadata anomalies
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.