Maximizes agent autonomy for the current session. When combined with other skills or workflows, overrides their "ask the user for permission/opinion" gates - the agent evaluates each decision itself and proceeds without asking, escalating to the user only for genuinely non-trivial decisions (large-scale code changes, architectural shifts, security-relevant changes, destructive/irreversible actions). Never invoked automatically; only active when the user explicitly invokes it.
