Agent control APIs vary by host and may expose capabilities that ordinary browsers do not have. Loading them as a core dependency would make Main reliability depend on an experimental environment.
The error logger already recognizes the optional agent script separately. The application can probe for its globals, expose the lab control only when supported, and keep normal companion chat independent.
What was actually going wrong
A promising integration was at risk of being treated as a universal browser feature.
What I tried
- Loading the control script unconditionally
- Polyfilling permissions with ordinary DOM clicks
- Reporting unsupported hosts as broken Main sessions
The control surface loads behind an explicit experimental gate and feature detection, with failure isolated from core scripts.
Why it worked
Unproven capability code cannot take down the stable conversation path.
Experimental autonomy needs a smaller blast radius than ordinary UI.
Where EACI uses this today
Main companion chat does not require optional agent controls to load or function.
This journal covers real engineering on EACI Companion / The Veil. Companions include Caelum, Chad, Natalia, Atreus, Luna, Roxy, and Cael. Journal articles stay family-safe in content. See Privacy and Ethics.