mind-security

What This Skill Does

The mind-security skill is a powerful defense mechanism integrated into OpenClaw for verifying the authenticity of digital media. By leveraging Bittensor Subnet 34, this tool provides sophisticated, adversarial-trained detection capabilities to identify deepfakes and AI-generated images, videos, and audio. Unlike simple pattern matching, this skill uses a multi-layered verification pipeline including C2PA metadata analysis, similarity matching against known AI datasets, and advanced 'absurdity' analysis to detect physically impossible events in videos.

Installation

To add this capability to your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/canvinus/mind-security

Ensure you have a valid BITMIND_API_KEY. You can obtain this by registering at app.bitmind.ai and generating a new key from your dashboard. Once installed, set your environment variable to allow the scripts to authorize requests automatically.

Use Cases

• Journalistic Verification: Quickly vet viral media or social media claims before reporting or sharing.
• Content Moderation: Automate the screening of user-uploaded images and videos to filter out deceptive AI-generated material.
• Personal Safety: Check videos of public figures or suspicious communications to identify synthetic manipulation intended to spread misinformation.
• Asset Integrity: Verify the provenance of images used in professional design workflows.

Example Prompts

1. "OpenClaw, please analyze this image [link] and tell me if it is AI-generated. Include the confidence score."
2. "I suspect this viral video is a deepfake. Run a security check on this link: [link] and provide the absurdity analysis if available."
3. "Check if this social media post image has any C2PA metadata or signs of AI manipulation. Use the mind-security tool."

Tips & Limitations

• Deep Analysis: Use the debug: true flag when checking videos to receive raw scores, processing times, and detailed absurdity analysis, which explains exactly what the model detected.
• Chain of Evidence: The isAI logic is cumulative; if C2PA data confirms the file is synthetic, the confidence level is immediate. If not, the system proceeds through similarity and visual analysis.
• Limitations: No detection tool is 100% accurate. AI generation is evolving rapidly, and 'adversarial competition' means the model is constantly catching up to new generators. Always treat the result as a probability score rather than an absolute fact. For critical investigations, corroborate with other forensic methods.