executing-planslisted

# Executing Plans ## Overview Load the plan, review it critically, execute the tasks in order, and hand the finished branch off for review. You are the single writer on an isolated branch or worktree; never start implementation on a main or master branch without explicit user consent. **When this skill vs subagent-driven-development:** one criterion, stated the same way in writing-plans and SDD. Prefer megapowers:subagent-driven-development when subagents are available, the plan's tasks are mostly independent, and per-task commits are acceptable; review quality is significantly higher with fresh per-task subagents. Use this skill when subagents are unavailable, when tasks are tightly coupled, or when you or your human partner want inline single-writer execution with your own commit cadence. ## The Process Read the plan and review it critically before executing anything. Raise concerns with your human partner first. Under an active autonomous-run charter at level `autonomous` or `on-the-loop`, resolve non-blocking concerns yourself, journal each resolution, and proceed; stop only for a real blocker. With no open concerns, declare one progress surface, the plan checkboxes or a ledger, and start. Do not duplicate the plan into conversational todos. Execute tasks top to bottom. The plan is the specification: follow its steps, run its verifications as written, and invoke any skills it names. A task is complete only after its verification passes, and its completion is persist