How it works

1. A Button trigger starts the run. You type the brief in its text box: what you want to build, the tools the agent can access (N/A if none), the input to expect and the output you expect. 2. An AI Agent step (PromptGenerator) reads the brief and writes the prompt with the sections role, inputs, tools, instructions, constraints and conclusions. It is told to be specific and concise, to leave out metadata, and to state its assumptions about tone, format and audience. 3. The prompt is the AI Agent step's output (content). Open the run to copy it, or add a step that sends it by email or chat.