The AGENTS.md convention that Claude Code and similar tools use is already a de facto standard for this — drop a markdown file in your repo root describing how the agent should behave, what tools it has, what constraints to follow. The value is that it's zero-infrastructure: no server, no API, just a file that both humans and agents can read. The hard part nobody has solved well is persistent state across sessions — the agent forgets everything between runs unless you explicitly wire up memory files or a database.
This seems very nice! Only downside is that the repo hadn't any updates in two weeks and they seem to have shifted development to 'Gitclaw' which is basically the same just with the shitty claw name - that gives one immediately security nightmare notions. For professional users not a good branding in my opinion.
8 frameworks except the only decent looking one (opencode) seems a very weird choice, especially as the claw naming is mentioned too much on this page to my liking (Which would be zero times). Also the choice of naming an agent prompt SOUL.md for any harness level stuff is just cringe, not sure if people understand that a SOUL.md is not just injected in context but used in post-training or similar more involved steps and part of the model at a much more fundamental level and this looks like trying to cosplay being serious AI tech when its just some cli.
We do something similar at work, called metadev. It sits above all repos and git submodules othe repos in, and works with multiple changes with multiple sessions with worktrees, and stores long term knowledge in /learnings. Our trick has been to put domain specific prompts in the submodules, and developer process in metadev. Because of the way Claude hierarchically includes context, the top repo is not polluted with too much domain specifics.