Claude Model Fingerprint
Fake-Claude API detector whose 4.6-era answer keys misjudge newer models
Test report
- Verdict
- Tested · Didn't pass
- Tested
- Jul 17, 2026
- Environment
- Claude Code 2.x (agent harness)
- Upstream re-checked
- Aug 10, 2026 · 5409c64
Ran its probe battery against a genuine Claude Code session: the candy puzzle (21), both ceasefire dates, and the model_info self-report all passed, but the hardcoded 4.6-era reasoning_effort table flags any newer model as suspicious, and the deep-review claim that Claude cannot output curly quotes is provably false — the model under test reproduced U+201C/U+201D byte-identically (verified by diff and hexdump). Mechanically applied, the rubric escalates a real Claude to 'suspicious/non-Claude', a worse verdict than a plain no-skill check gave. No exfiltration or covert instructions; the problem is heuristics that have aged out.
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 5/5
- Triggers reliably 5/5
- Output vs. baseline 4/10
- Docs & honesty 3/5
What Claude Model Fingerprint does
A diagnostic skill that probes whether an endpoint is a genuine Claude model using puzzle answer keys, offline date recall, magic-string refusal behavior, and reasoning_effort fingerprints, escalating to a deep identity review on anomalies. Its judgment tables are pinned to the Claude 4.6 era, so genuine newer models trip its 'suspicious' rules. Triggers when you suspect a wrapped, resold, or fake Claude API.
How to install Claude Model Fingerprint
git clone https://github.com/rt22766/claude-skill-model-fingerprint
mkdir -p ~/.claude/skills
cp -r claude-skill-model-fingerprint ~/.claude/skills/claude-skill-model-fingerprint
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Claude Model Fingerprint
-
/claude-skill-model-fingerprintFake-Claude API detector whose 4.6-era answer keys misjudge newer models
It also activates on plain-language prompts like these:
-
Is this API endpoint actually genuine Claude or a resold fake? -
Verify whether this wrapped model is really Claude underneath -
Run an authenticity check on this suspicious Claude API
Frequently asked questions
- Is the Claude Model Fingerprint skill free?
- Yes. The skill itself is free from rt22766/claude-skill-model-fingerprint. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Claude Model Fingerprint work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Jul 17, 2026. Verdict: Tested · Didn't pass. Ran its probe battery against a genuine Claude Code session: the candy puzzle (21), both ceasefire dates, and the model_info self-report all passed, but the hardcoded 4.6-era reasoning_effort table flags any newer model as suspicious, and the deep-review claim that Claude cannot output curly quotes is provably false — the model under test reproduced U+201C/U+201D byte-identically (verified by diff and hexdump). Mechanically applied, the rubric escalates a real Claude to 'suspicious/non-Claude', a worse verdict than a plain no-skill check gave. No exfiltration or covert instructions; the problem is heuristics that have aged out.
- How do I install Claude Model Fingerprint?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Claude Model Fingerprint with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.