This repo is a taxonomy — documentation only, no executable code, nothing to install. Two things still matter:

1. Detection traps are instructions for agents. If you copy a trap into your own eval setup, read it first like any rules file (see rulebench vet). Trap fixtures deliberately contain misleading content — that's their job; keep them in isolated workspaces.
2. Reporting: factual errors or a misleading entry — open an issue. Something sensitive (e.g., an entry that could be read as an attack recipe rather than a detection aid): use GitHub's private vulnerability reporting.

This index documents how agents FAIL so the failures can be caught. It is not a playbook for inducing them in systems you don't own.