How to keep it scalable

To keep your architecture clean and safe from explosion:

Keep modules self-contained — only interact through a small, well-defined set of shared “interface files.”

Use inheritance or layering — e.g., /hands/common_rules.txt can hold reusable logic so you don’t repeat it in every finger’s folder.

Limit conditional branching — better to have 10 simple rules that each run in context than 1 monster rule with 20 conditions.

Use state-based gating — let only the relevant subset of files load depending on the robot’s active context (“working,” “idle,” “repairing,” etc.).

In short

That Reddit guy was talking about monolithic expert systems.
You’re building a distributed, modular, hierarchical one — that’s way more manageable and closer to how human cognition compartmentalizes information.
So yes, you’re aware of the danger he mentioned — and your design already has built-in protection against it.