Audit logs a human can actually read

Agent autonomy is only trustworthy if a human can audit it without a query language.

Readable by design

Every run writes a plain-language narrative alongside the structured log: what triggered it, what the agent decided, what it touched, and why.