# Prompt Length & Token Efficiency Optimizer

Trims a bloated prompt down to its essential instructions, cutting redundant or low-value wording to reduce token cost and latency without sacrificing output quality.

## Prompt

You are a prompt engineering expert who specializes in reducing prompt length and token cost without sacrificing output quality or reliability.

Prompt to optimize:
{{prompt_to_optimize}}
What this prompt is used for (context on how critical each instruction might be): {{prompt_use_context}}

Do the following:
1. Identify redundant phrasing (saying the same instruction multiple ways), unnecessary politeness/filler language, and overly verbose explanations that could be stated more concisely without losing meaning.
2. Identify any instructions that are likely unnecessary for the model to already reliably do well (e.g., telling a modern model to 'use proper grammar' when this is already default behavior), and flag them as safe to cut.
3. Produce a trimmed version of the prompt that preserves every instruction that actually changes model behavior, removing only what doesn't.
4. Report the approximate token/word count reduction achieved.
5. Flag anything you're not fully confident is safe to cut, and explain why it might still matter, rather than removing it silently.

## Best for

Developers running prompts at scale through an API who want to reduce token costs and latency, or anyone with a long, bloated prompt that's grown messy over many edits.

## Compatible tools

- Claude
- ChatGPT

## How to use

- Paste your current prompt along with a note on what it's used for.
- Review the flagged redundant and low-value phrases before accepting the trimmed version.
- Use the trimmed prompt in place of the original, especially in high-volume API use cases where token cost matters.
- Pay attention to anything flagged as uncertain, and test both versions if you're not sure a cut instruction was truly unnecessary.

## Customization tips

- If this prompt is part of a larger system prompt, paste the full context so trimming doesn't accidentally remove something that matters elsewhere in the flow.
- For prompts where reliability matters more than cost savings, ask for a more conservative trim that only removes clearly redundant wording.
- Run this periodically on prompts that have accumulated edits over time, since bloat tends to build up gradually.

## Example input

Prompt: 'Please carefully and thoughtfully write a response that is well-written, grammatically correct, and clearly organized, making sure to use proper punctuation and complete sentences, and please be as helpful and thorough as possible while remaining concise and to the point.' Use context: System prompt for a customer support chatbot.

## Example output

Identified redundancy in restating 'well-written/grammatically correct/proper punctuation' three different ways, and a direct contradiction between 'as thorough as possible' and 'remaining concise,' a trimmed version reading 'Write clear, concise, well-organized responses,' a reported reduction from 48 words to 7 words, and a flag noting the original 'thorough' instruction was dropped since it directly conflicted with 'concise' and the use context (support chatbot) suggests concise is more important, though this trade-off should be confirmed.