dsh-prompt-polish

Writing prompts in DeepSeek Harness often means wrestling with vague drafts, missing constraints, and repetitive rewrites. dsh-prompt-polish adds an Optimize button right in the composer, sending your current draft through the model provider already configured in DSH so you never have to open another tool or hand-craft a meta-prompt. The result appears in a review popover where you compare the original and optimized text side by side before choosing to adopt, keep, or dismiss—nothing is ever written automatically.

Six rewriting modes cover the main use cases: Default (precise), Concise, Structured, Creative, Translate and optimize, and Code request. You can optionally inject session context—goal, todo list, compaction summary, recent messages, and tool-result snippets—so the optimized prompt reflects the live task. Language follows your draft, locks to Chinese, or locks to English, and a 500-character custom-instruction field lets you fine-tune behavior without breaking the output contract.

If your daily workflow in DeepSeek Harness involves drafting multi-step instructions, describing code requirements, or rephrasing Chinese thoughts into clean English prompts, this plugin saves you from the repetitive cycle of manual polishing and keeps your focus on the task itself.