Concise instruction rewriter
Prompt instructions get bloated. We write to models the way we write to people — “I would really like you to please go ahead and carefully…” — and all that padding adds tokens without adding signal. Models follow tight, imperative directives more reliably. This rewriter mechanically strips the hedges, polite preambles, and wordy phrases so your instructions say exactly what they mean and nothing more.
How it works
The tool applies a set of deterministic compression rules to your text. It deletes polite preambles such as “I would like you to please,” removes hedge and filler words like “really,” “just,” “basically,” and “carefully,” and collapses common wordy phrases — “in order to” becomes “to,” “due to the fact that” becomes “because,” “the majority of” becomes “most.” It then tidies spacing and capitalization and reports how many words were saved. Everything runs locally, so it is instant and private.
Tips and notes
- Re-read the output. Mechanical compression is conservative but not infallible; confirm no essential constraint was trimmed.
- Use it on instructions, not user copy. Imperative terseness is great for system prompts and bad for a friendly chatbot greeting.
- Shorter means cheaper. Fewer tokens in the instruction means lower cost and latency on every call that uses the prompt.
- Combine with a scope check. After tightening, run the variable scope checker to make sure no placeholder got mangled.