distillation-implementationlisted

# Distillation Implementation (Stage 3) Execute the distillation plan, task by task. Each task carries its mode, its keep-verbatim items, and its seam substitutions — bring the reference's encoded decisions into your project under that mode, preserving the keep-verbatim items and wiring the seams to your dependencies. **Why subagents:** You delegate logic-heavy chunks to fresh agents with isolated context. By crafting their instructions precisely — the task's mode, its keep-verbatim items, its seam substitutions — you keep them focused and stop the reference's packaging from leaking into your project. They never inherit your session's history; you hand them exactly the chunk they need. This also preserves your own context for coordination work. **Core principle:** Fresh subagent per logic-heavy chunk + two-stage distillation-aware review (spec compliance then code quality) = faithful port, fast iteration. Simple copies go direct. **Continuous execution:** Do not pause to check in with the user between tasks. The human gates are between stages, not between tasks. Execute the whole plan, then finish the branch. Stop only for a BLOCKED you cannot resolve, genuine ambiguity, or completion. ## When to Use You're in Stage 3: the distillation plan is approved. Execute it. Two decisions drive each task — whether it takes the subagent path or the direct path, and (on the subagent path) what mode it runs under. ```dot digraph when { "Task: logic-heavy or risky?" [shape=diamo