Write a ticket with limits before the agent starts
Ming Chen ran his blog for two weeks almost entirely through agent sessions. Every change starts as a GitHub issue with context, scope, what is out of scope, acceptance criteria and one risk label, and separate sessions review every pull request. He writes that agents readily deliver something that looks finished without being so, and do more than the issue asked for. He says the setup suits small, reversible tasks.
Done with Valgfri agentModerate
What was done
Ming Chen ran his blog for two weeks almost entirely through agent sessions. Every change starts as a GitHub issue with context, scope, what is out of scope, acceptance criteria and one risk label, and separate sessions review every pull request. He writes that agents readily deliver something that looks finished without being so, and do more than the issue asked for. He says the setup suits small, reversible tasks.
How to do it
- Paste the prompt into the tool, and replace the text in brackets with your own.
- Read the answer through. Do not send, delete or publish anything until you have seen it yourself.
- Check the original before you trust a figure, a name or a date.
Try it with Valgfri agent
Write a GitHub issue for [change] in [repo]. Include background, what is to be done, what is explicitly out of scope, acceptance criteria I can check and one risk label: [low, medium or high]. Only one change per issue. Do not open any pull request and do not change code. You are done when the issue is so precise that another session can do it without asking me.
Replace text in brackets with your own.
Ming Chen, blog ·