System prompt length optimizer
Every token in your system prompt is paid for on every single request, so a bloated prompt is a recurring tax. This tool reads a system prompt and flags the sentences most likely to be removable — filler phrases, repeated constraints, and run-on instructions — and estimates the tokens you would save by cutting them.
How it works
The prompt is split into sentences and each is scored on three signals. First, filler: a dictionary of low-information phrases like “please note that” and “it is important to” marks padding. Second, redundancy: each sentence is reduced to a word-bigram set and compared with earlier sentences using the Jaccard index, so a constraint restated later gets flagged. Third, length: unusually long sentences are flagged as compression candidates. The overall verbosity score combines these with average sentence length, and a rough four-characters-per-token estimate shows the total and the potential saving from removing every flagged sentence.
Tips and notes
- Cut filler first. Politeness padding and meta-instructions (“note that”, “keep in mind”) almost always delete cleanly with no behavior change.
- De-duplicate constraints. Stating the same rule twice does not make the model follow it harder; keep the single clearest phrasing.
- Prefer lists over prose. Converting long instruction paragraphs to terse bullet points usually beats deleting them outright.
- Re-measure after each edit. Paste the revised prompt back to confirm the token count actually dropped before you ship it.