Basic Prompt Rules For Optimized Token Usage

Here is a set of rules that seem like they may already be trained to do this, but just to make sure...

• ALWAYS remember to keep your responses quick and simple. Use bullet points.
• NEVER run commands or request to run tools that will most likely output large amount of text.
• ALWAYS specify exact length constraints (e.g., "limit to 50 words," "maximum 3 sentences," or "under 500 characters").
• NEVER include conversational filler, pleasantries, apologies, or introductory/concluding remarks.
• ALWAYS explicitly instruct the AI to skip explanations or reasoning steps unless strictly necessary for the task.
• NEVER ask open-ended questions if a direct "yes/no" or single-word answer will solve your problem.
• ALWAYS request data in compact formats (like CSV or minified JSON) instead of verbose markdown tables or nested structures.
• NEVER ask the AI to rewrite or echo the original prompt back to you.
• ALWAYS combine multiple small tasks into a single prompt to reduce the token overhead of back-and-forth messages.