dsh-simple-view

In a long DeepSeek Harness session, the transcript gets cluttered fast: Think reasoning rows, context-injection/recall lines, and tool-call entries stack up between every real reply, and the generous spacing and large fonts make the screen feel crowded. dsh-simple-view is a focused cleanup that lets you see only what matters.

It hides execution-log rows from the transcript while keeping error lines and ask_user_question prompts fully visible, so failures and follow-up questions are never lost. It tightens the gap between message blocks, removes dead space under each one, shrinks the font size for markdown body text and headings, and wraps assistant replies in a crisp 8 px square-corner bubble that matches your own message style. The most flexible piece is an editable instruction card in Settings → Plugins → dsh-simple-view; the text you write there is injected into every request's system prompt, nudging the model to skip process narration and lead straight to the conclusion. Leave it blank and the injection turns off entirely.

If you run long, tool-heavy conversations in DSH and get frustrated by dozens of execution lines drowning out the actual answer, dsh-simple-view narrows the view down to question, conclusion, and error—lighter, cleaner, and easier to scan.