Agent Rule Audit

Audit an OpenClaw agent's behavior-layer rules and prompt sources to find drift, redundancy, conflict, loss of focus, and weak behavior guidance. Use when reviewing an agent's core behavior files, any behavior-shaping correction/learnings layer that exists in the target workspace, and any explicitly referenced shared rule files; when checking why an agent behaves poorly despite many rules; when deciding which rules should stay core, move out, or be strengthened; and when running an agent self-audit.

by @gkso (Alan Wang)·MIT-0