dsh-prompt-optimizer

Prompts typed in a hurry are often vague and unstructured, and the model responses follow suit—sometimes sharp, sometimes off. dsh-prompt-optimizer addresses this everyday friction by adding a single button right below the chat composer in DeepSeek Harness, so you can polish a rough draft into a clear, well-organized prompt without ever leaving the conversation.

Click the button and the plugin reads whatever is in the input box, then walks the host model through a fixed six-step rewrite: clarify intent, surface context, expose hidden constraints, identify gaps, restructure the prose, and run a language-consistency check. Code blocks, technical terms, and URLs are preserved throughout, and the output always matches the language of your input. A side-by-side before-and-after dialog appears when the rewrite is ready; one click on "Replace into input" swaps the polished version back into the composer so you can send it immediately.

It is built for two audiences: developers and engineering teams who use dsh daily and want fewer rounds of trial-and-error, and technical writers or prompt-crafting enthusiasts who want more consistent output from the same model across different tasks. As long as your dsh profile has any working model configured, installing this plugin gives you a lightweight quality checkpoint on every single prompt.