Skill · Prompt

Run to the gate, then stop — not step-by-step

A prompt that lets the agent run on its own but stop at the few places worth stopping — so you can let go without losing control.

TypePrompt

LanguageĐọc bản tiếng Việt →

You hand the agent a long task, then sit there watching it type line by line, finger over the cancel button. Supervising like that, where's the saving? But last time you let it run all the way to the end, it sailed past a wrong decision in the middle and you found out forty minutes later. So you're stuck between two options that both feel wrong.

There's a third way, and it isn't about whether you trust it — it's about putting the gate in the right place. You don't need to see every step. You just need to be asked at the few moments where a wrong step is expensive: before it does something hard to undo, before it switches phases, and right when it's unsure. Everything else — let it run.

✕ Binary

Watch every step → slow, draining, no saving left

▼

…or let it run free → it overshoots, you find out too late

▼

Either choice feels wrong

✓ Run to the gate

Runs the safe parts on its own, doesn't bother you

▼

Stops at the gates: destructive · phase change · unsure

▼

You let go and still hold the moments that matter

Paste it when you hand off a multi-step task you want to walk away from but aren't ready to leave entirely. The trick is naming the gates up front, so it knows when it may decide on its own and when it must look up and ask — instead of asking about everything (annoying) or asking about nothing (risky).

Paste into the chat

Work through this task autonomously — don't ask me about routine steps. But treat these as hard stops: pause and wait for my go-ahead before each.

- Before anything that writes, deletes, deploys, sends, or is otherwise hard to undo.
- Before you move from one phase of the work to the next.
- Whenever you hit a decision you're less than about 80% sure of.

At each stop, give me three short lines — what you just did, what's next, and why — then wait. Everything between the gates, just do.

Prefer your agent install it?

Add this rule to my CLAUDE.md, verbatim:

Work through this task autonomously — don't ask me about routine steps. But treat these as hard stops: pause and wait for my go-ahead before each.

- Before anything that writes, deletes, deploys, sends, or is otherwise hard to undo.
- Before you move from one phase of the work to the next.
- Whenever you hit a decision you're less than about 80% sure of.

At each stop, give me three short lines — what you just did, what's next, and why — then wait. Everything between the gates, just do.

Did this help?

Tags#autonomy #skill

Other skills

01Verify Before You Call It Done 02Plan First, Then Code 03Map the Blast Radius Before You Edit 04If It's Deterministic, Don't Make It Guess 05Diagnose From the Symptom, Not the Guess 06Make It Argue Against Its Own Plan 07Hand Off to Your Next Session 08A Review That Doesn't Rubber-Stamp 10A CLAUDE.md skeleton — turn a description into operating law 11The weekly review prompt — five minutes that turn experience into skill 12Split the work into focused slices, each with clean context 13Is this even worth an agent — ask before you open the chat 14Probe with a small task before you trust it in a new domain

Back to skills

The author

Each story here wraps a lesson paid for in full.

craftagentsomeone building and learning at once

What are you building with agents? Want to trade notes, push back, or build something together — drop a line.

Write hello@craftagent.cloud

58pieces12clustersVI·ENbilingual

✕ Binary

✓ Run to the gate

Get new pieces by email