THE CRAFT · HOW TO DO IT

Prompting Agents

Prompting a chatbot and prompting something that acts on your machine are different skills. This is the method for the second one.

◈ IN PREPARATION

This page is scaffolded and its content is being written. What it will cover: Scope, stopping conditions, verification steps, and why an agent that cannot stop is the most common failure mode.

Every method here is dated. An agent technique that cannot survive a model release was never a method.

Meanwhile

The evidence underpinning this section is already published:

What agents actually do — measured success rates, the benchmark that saturated, and the security findings.
How to ask AI well — the habits that survive a model release.
When not to use AI — the nine cases where reaching for it costs you.