claude-skill-model-fingerprint

What problem does it solve?

When using Claude Code or API integrations, it is difficult to confirm whether the underlying model is a genuine Claude instance or a third-party wrapper, downgraded version, or proxy. This Skill solves that problem by providing a systematic, multi-signal verification workflow.

Core Features & Use Cases

• Multi-signal fingerprinting: Combines reasoning_effort detection, Magic String refusal tests, combinatorial reasoning puzzles, and offline date memory checks to cross-validate model identity.
• Parallel subagent execution: Runs core tests simultaneously via three subagents for rapid results without sequential bottlenecks.
• Conditional deep review: Automatically escalates to comprehensive identity, tool ecosystem, metadata, and nested-layer analysis when core tests reveal anomalies.
• Use Case: A developer suspects their API endpoint is routing to a third-party wrapper instead of real Claude. Use this Skill to run a quick authenticity check and generate a structured diagnostic report.

Quick Start

Use the claude-skill-model-fingerprint skill to verify whether the current model is a genuine Claude instance and generate a diagnostic report.