Question
What may the primary controller do autonomously after one task-level authorization, and exactly which boundaries require human escalation?
Resolve the policy envelope for repository mutation, network and external effects, model/API cost, skill/tool classes, recursive depth, stop conditions, and risk tiers. Preserve long-task continuity with local branch freezing and global continuation, while never treating silence or stale evidence as approval.
Question
What may the primary controller do autonomously after one task-level authorization, and exactly which boundaries require human escalation?
Resolve the policy envelope for repository mutation, network and external effects, model/API cost, skill/tool classes, recursive depth, stop conditions, and risk tiers. Preserve long-task continuity with local branch freezing and global continuation, while never treating silence or stale evidence as approval.