A skill for identifying and verifying AI model identity through multi-stage behavioral analysis. Works with omp, Claude Code, and any agent that discovers ~/.agents/skills/.

• Multi-stage identity verification protocol
• Knowledge boundary probes (time-anchored events)
• Behavioral fingerprinting (stylistic patterns, refusal vectors)
• Statistical pattern analysis
• Quick fingerprint tests for rapid identification

bunx skills add ParticleG/llm-fingerprint-skill

Or manually:

git clone https://github.com/ParticleG/llm-fingerprint-skill ~/.agents/skills/fingerprint

Trigger phrases:

• "identify this model"
• "fingerprint the AI"
• "verify model identity"
• "knowledge cutoff test"
• "who are you really"

This skill incorporates techniques from recent academic papers:

• Hide and Seek (arXiv:2408.02871)
• RoFL (arXiv:2505.12682)
• TRAP (arXiv:2402.12991)
• UTF (arXiv:2410.12318)
• And more...

MIT