orchestrating-executionlisted

# Orchestrating Execution Execute a plan by dispatching a **fresh subagent per task**, gating each finished task through **exloom's review** (tier-scaled), looping fixes until clean, and running one **whole-branch review** at the end. You — the controller — never write the code; you coordinate, curate context, and hold the line on the gate. This is the high-throughput execution mode. A solo developer gets a task-by-task review panel they wouldn't otherwise have; a team gets the same with the handoff discipline enforced. It pairs with `exloom:planning-for-handoff` (which produces the plan) and `exloom:review-gate` (which defines the tiers and the reviewers). ## When to use it - You have a written plan with discrete tasks (`exloom:planning-for-handoff`). - You want to stay in this session (no parallel-session handoff). - Tasks are mostly independent or have a clear order. If the plan doesn't exist yet, go to `planning-for-handoff` first. If a task needs design work the plan didn't anticipate, stop and go to `brainstorming` — don't improvise it inside an implementer. ## Why a fresh subagent per task Each implementer is constructed with exactly what it needs — its task, the interfaces it touches, nothing else. It never inherits your session history. This keeps each implementer focused and keeps **your** context clean for coordination. The cost of a task lives in its own subagent, not in your context window. ## Workspace setup (once) Before the first dispatch, isolate the