Claude Code's goal command takes a finish line and keeps working until it crosses it. These 5 lines write a finish line Claude can actually check.
The goal command changes how Claude Code works. Instead of you feeding it one instruction at a time, you describe the finished result and Claude plans, works, checks itself, and keeps going until it gets there.
Most people write their goal as one vague sentence, so Claude wanders. The template below writes the goal in 5 parts, and the last part is the one everybody skips: how Claude proves it did the job. It has been saving me hours every single week.
Fill in the 5 lines, then hand the whole thing to the goal command. Done means what you wrote, and Claude has to prove it.
TASK: [What Claude should do] WHY: [Why this matters or who it is for] OUTCOME: [The exact finished result] CONSTRAINTS: [What Claude must or must not do] VERIFICATION: [How Claude should prove the outcome is complete]
This is the exact example from the video, a week of content in one goal.
TASK: Install this plugin to create a week of content. WHY: So I show up every day without writing from scratch. OUTCOME: A week of ready to post content for different platforms. CONSTRAINTS: Match my voice file and fit each platform's limits. Stop after 10 turns. VERIFICATION: Check every post against my voice rules and keep fixing until each one passes.
The goal command lives in Claude Code. If you haven't seen it yet, Anthropic's official guide shows what it can do.