Motivation: a lot of agent issues I’ve seen aren’t model problems—they’re missing-process problems (unclear scope, inconsistent tool use, no operational playbook, no shared constraints). The goal here is to make that “boring but critical” layer easy to reuse so each new agent project doesn’t start from scratch.

Repo: https://github.com/thompson0012/agents-docs-kits/tree/main

I’d love feedback on:

What doc artifacts you’ve found most valuable for agents (beyond the prompt itself).

How you structure “agent instructions” / repo-level guidance so it stays maintainable.

PRs with templates: eval checklists, tool-safety policies, incident/runbook patterns, or onboarding docs.

What would you want in a “starter docs kit” if you were spinning up a new agent repo today?