← All guides

A copy-desk pass in your browser — not a cloud rewrite.

The checklist behind PromptOptBase rewrites: task, constraints, format, filler cuts, and what we never invent.

Last updated: 2026-07-13

What runs locally

When you click Optimize, your draft is transformed by deterministic rules in the tab. No prompt is uploaded for generation, scoring, or storage. The badge Structured rewrite · no model API call is literal.

If you want a model to critique the prompt, paste the output into ChatGPT, Claude, or Gemini yourself — we do not broker that call.

The checklist

We look for four common failure modes: buried task, missing audience or tone, unnamed output format, and filler phrases that eat context window. Each fix shows as a redline in the before/after view.

Model presets only change labels and phrasing habits. They do not change the checklist or call vendor APIs.

What we refuse to invent

The optimizer does not add product metrics, customer names, or policy guarantees you did not supply. Expand mode on the Length tuner adds empty labeled slots — not fabricated claims.

You are always the editor. We structure; you judge facts, tone, and whether to send.

Go deeper

Methodology lists the same rules with tool-by-tool notes. Framework and marketing tools use separate assembly templates but the same no-API discipline.

Structured rewrite · no model API call