Notes on handing work to coding agents
Working notes, updated as I learn things the hard way.

Why these notes exist

Most teams now have more agent capacity than they have well-written tasks to spend it on. These are notes on closing that gap.

Over the past year the bottleneck on my team stopped being “can an agent do this?” and became “can we describe this well enough that an agent gets it right on the first pass, and a human can check the result?” Those turn out to be two different writing problems, and most task templates only solve one.

These pages are the parts I keep having to re-explain. They are deliberately short and opinionated. Where someone else has written the idea up properly, I link out rather than restate it.

Pages