Agents parrot explanatory text from CLI output

If a CLI built for agents returns a human-readable explanation in its JSON (e.g. an info field explaining background behavior), the agent tends to paraphrase it into its final reply to the user, creating noise. Keep agent-facing command output minimal (e.g. {"saved": true}) and put the why in an instructions file the agent reads once. Removing the sentence from output, not adding a do-not-mention rule, is the reliable fix.