Model-facing config patch/apply calls cannot newly enable dangerous security-audit flags.
Gateway-tool config patch guardrails
Model-facing gateway tool calls now reject config.patch and config.apply when a patch would newly enable any flag enumerated by openclaw security audit.
Scope
- Direct authenticated operator RPC behavior is unchanged
- Already-enabled dangerous flags are not retroactively blocked in unrelated edits
- New dangerous toggles are blocked at model-facing tool surfaces