Read CLAUDE.md — it is the single source of truth for this repository.
CLAUDE.md is the canonical guide for every agent (Claude, Codex, or otherwise). It
covers the architecture (parse → plan → apply → run), the module map, the
.managed-agents/ folder convention, per-provider status (Anthropic / Google /
OpenAI), the commands, the dev workflow, and the ground rules — including the
mandatory secrets/anonymization rule to follow before every commit.
This file intentionally adds no rules of its own; follow CLAUDE.md.