A module cannot bypass core

The user’s rule of 2026-10-08 (BRIEF.md, “CORE CANNOT BE BYPASSED”), in the user’s words: “its very important that a module cannot bypass core modules, its not about making a module that does not bypass, but make the harness so a malicious module just cannot bypass core”.

• Core is logging, the content guard and redaction, outbound HTTP, the filesystem, process spawn, the database, crypto and keys, authorization and configuration.
• Assume the module is MALICIOUS: static guards and the architecture manifest are early feedback, never the boundary.
• Enforcement is by a boundary the module cannot touch: a process boundary with kernel confinement plus a capability channel to core, and/or frozen-realm isolation.
• Every design and review states how a malicious module is stopped.

It is carried out by CORE-NO-BYPASS.