Designing a fair benchmark for local security agents
A practical test plan for comparing local models on instruction following, tools, scope control and repeatability.
THE ARCHIVE
Transmissions from the lab.
Study notes say what is still theoretical. Field logs document the method. Investigations make evidence explicit.
01
FIELD LOG 004 · build log
02AI RED TEAMING NOTE 003 · study note
03What transformer attention changes about AI security
A study note on context, token relationships and why model understanding is not the same as a security boundary.
BUILD LOG 002 · field note
Where autonomous security testing should stop
A boundary map for automation, evidence, manual verification and accountable reporting.