dsh-prompt-boost-pro

Writing prompts in DeepSeek Harness often means fighting with rough drafts - vague goals, missing constraints, no expected output format. Send one as-is and the model can only guess. dsh-prompt-boost-pro adds a sparkle button to the composer tool row: click it, pick a mode, and the model rewrites your draft into a sharper instruction. The result appears as a side-by-side original-versus-enhanced preview; you can accept, re-run in the other mode, or discard. Your draft stays untouched until you say yes.

Two modes cover different needs. Structured reorganises the draft around goal, constraints, expected output, and acceptance criteria - ideal for task-style prompts. Light only disambiguates, fixes wording, and tightens phrasing - perfect for drafts that are shaped but need polish. Neither invents requirements you did not write; at most one placeholder survives, and only for a decision that is genuinely yours. The comparison preview, an undo button that expires the moment the draft changes, an adaptive output budget that flags truncation instead of dropping a capped answer, and a four-level model route resolution are the small details that make the whole flow feel reliable.

Built for developers who write complex prompts in DSH regularly and want to go from rough idea to executable instruction in one click without losing ownership of the final text. No extra configuration is needed - a brand-new session works on the first click, the UI is bilingual, and the plugin writes no config files and collects no telemetry.