Detection-Fidelity Score 3.0 blocks prompt injection

This looks more like a high-concept security showcase than a validated breakthrough: the pitch is heavy on exotic terminology, the benchmark lacks reproducible detail, and the missing repo blocks real peer review.

• –Safety DB lists `detection-fidelity-score` as a PyPI package with the description "A hardware-bound security layer for AI Agents using Tesla 3-6-9 logic," which at least confirms a package artifact exists.
• –The headline claim of "0 false positives in 10k tests" is unusually strong for prompt-injection defense, and without a published dataset, attack mix, or methodology it reads as marketing rather than proof.
• –Hardware-bound entropy could help with replay resistance on a single device, but it does not solve the broader prompt-injection problem of hostile instructions entering through tools, retrieved content, or multi-agent workflows.
• –The broken GitHub link is a major credibility hit for a project explicitly asking the community to inspect and break the system.