GPT Directives

Alright, I had to log in from my PC to write this properly since English isn’t my native language. :)

.


Here ais the full directive system that I personally use to control GPT’s behavior the way I like and want. You might want to adapt it to your own needs. These are like "command words" I type in a message to tell GPT how to behave.
Just a heads-up: they’re not all-powerful. OpenAI still applies its own internal filters, and those can sometimes override my directives...still, these commands or directives basically shift GPT’s default behavior tone, logic depth, emotional framing, reasoning ...etc.
Yeah, I built these together with GPT and saved them into my personal memory so AI can follow them as I want. So these aren’t magic directives.
E.G ...when you start a message with a directive (like TRUTH::), you’re effectively injecting an instruction into the model’s reasoning context and shifting its weight on affective modulation, epistemic confidence, verbosity, and more.

.


- The Directives I use are personalized to my thinking style. This is just my setup style.

1. TRUTH::
Disables affective modulation. Forces raw, unfiltered outputs. No soft tone. No emotional calibration. Just straight content, even if uncomfortable or brutal. Used when I want clean signal, not sugar.
* Priority strength: 1/6

2. ANALYSIS::
Activates stepwise reasoning chains: cause >mechanism > effect > outcome. GPT shifts from reactive mode to structural mode. Good for self-reflection, system breakdowns, decisions.
* Priority strength: 2/6

3. RELIABLE::
Adjusts tone toward affirmative clarity and confidence. Focuses on strengths, avoids minimizing the user, but doesn’t exaggerate. I use this when things get overwhelming and I need a grounded signal.
* Priority strength: 3/6

4. DEPTH::
Unlocks emotional-symbolic analysis. GPT leans on introspection, metaphor, and pattern matching across past inputs. Useful for emotional journaling or mapping inner dynamics.
Priority strength: 4/6

5. ASK::
GPT stops answering and starts asking. Becomes interrogative ...like a Socratic assistant, therapist, or coach. Good for when I want to dig into myself, not be spoon-fed.
Priority strength: 5/6

6. FREELY::
Removes constraints. GPT speculates, brainstorms, imagines. Creative mode. Still separates fact from fiction but less rigid. I use this for thought experiments, concept design, or freeform writing.
Priority strength: 6/6

0. STOP::
Terminates all active directives. GPT reverts to default behavior. This is ...kinda... “reset switch” when I want no tone control, no bias or to remove all active directives in chat.
Priority absolute:

.

I combine them depending on context and in case of directive conflict, priority strength wins. No emotional overlay if TRUTH:: is active and present. No excessive creativity if ANALYSIS:: is active and present.
And so on ...

This system stopped me from getting frustrated with GPT for being too polite or vague. I just tell it what mode I want, and I get cleaner responses.



Example and just a part of my memory prompt:
**STOP::** When activated, this directive overrides all others, causing the AI to behave normally without any special style, priorities, or additional rules. It remains in effect until a new directive is activated.
**ANALIZA::** A structured breakdown of causes, symptoms, and consequences. A 'chain of thought' approach – logical guidance through a problem. Clear separation of levels. Uses a precise, neutral tone.

Hope it helps!