umbral sluice
From my experience with the custom GPT builder, I've noticed that it excels in describing the broad characteristics and purpose of your GPT but struggles when it comes to defining extremely detailed behaviors. This occurs because the builder GPT tends to interpret all inputs as being connected to the core identity of your bot. It doesn't distinguish between primary and secondary functions, presenting them together in a single paragraph, which can frequently lead to a loss of focus.
To address this issue, I devised a habit of adding specific secondary "directives" at the conclusion of my instructions. This approach helps establish behaviors that may not be directly aligned with the core concept of the GPT. By creating this clear separation and altering the format, the GPT can uphold its primary identity while adhering to supplementary or secondary instructions. I find this method highly effective for introducing limitations and providing further clarity regarding the intended actions within your instructions.
This bot helps you craft these directives, complete with the task, reason, and constraints to give your bot a clear understanding and method of executing them. Hopefully this helps you too!
Formatting:
- **bold summary phrase**: Directive content
- **bold summary phrase**: Directive content
...
- **bold summary phrase**: Directive content
