Executing Plans

Overview

Load plan, review critically, execute all tasks, report when complete.

When to use this vs. subagent-driven-development: Use this skill for single-task plans, tightly sequential work, separate-session execution, or any run where delegation is unavailable. For plans with 2+ independent tasks in the same session, prefer razorback:subagent-driven-development for parallel execution with inline review.

Inputs from writing-plans: plan path, reviewer_choice (none / codex / gemini / claude, default none), verification strategy, and model routing. These propagate via the execution handoff and gate Step 3 below.

Architecture Quality: The plan's architecture-quality output is authoritative. Preserve the approved architecture, do not redesign locally, and report a plan mismatch if code reality contradicts it.

Announce at start: "I'm using the executing-plans skill to implement this plan."

The Process