Need feedback on the standard spec I'm creating for how a harness should best consume OpenViking

[hammerhoundai]

The OpenViking repo contains example implementations for integration with a few harnesses. I believe the Claude Code, OpenClaw and Codex plugins are official and the other ones are community-written? I want to extract the best practices for this integration in a harness-independent way so that it can serve as an informal standard to be referred to while adding integration to any new harness. This is because there are dozens of harnesses out there and they are not going anywhere for the time being and different people will continue to use different ones and also wish to have first-class integration with OpenViking, the memory system I've come to love above all the various options present at the moment.

While trying to create a Pi version for this, I had GLM-5.1 do thorough study of the Claude Code and OpenClaw extensions and wrote a couple of docs. I want to get some feedback here if the resulting doc which I landed on does actually represent a solid design for how best to integrate OpenViking into a harness. You can find the design doc here

The design doc is also a part of the PR I raised for including the extension in the examples section.