Prompt optimizer for Claude
Rules traceable to official prompting guidance
Claude follows instructions literally and responds to structure: XML tags, data before the question, and the reason behind an ask all measurably change output quality. Anthropic's own docs put the gain from correct long-input ordering at up to 30%.
What the rewrite does
- Separates instructions from data with XML tags, the structure Anthropic's guides are built on
- Puts long material first and the question last (up to 30% quality gain on 20k+ token inputs, per Anthropic)
- Adds the motivation behind constraints, which Claude generalizes from
- Drops chain-of-thought scaffolding for thinking models, where it hurts rather than helps
Nothing is invented: missing details become clearly marked placeholders, and we don’t store your prompts. 10,000 free words a month.