dsh-writing-studio

Writing a 3000-word-plus article in a conversational AI is fragile: session timeouts, context compression, or simply switching machines can wipe out your progress mid-draft. dsh-writing-studio turns long-form writing into a trackable pipeline — create a project, set the outline, draft section by section, review each section, then assemble the final manuscript. Every step is persisted as a plain JSON file, so you can pick up exactly where you left off without relying on any single session's memory. The plugin has zero runtime dependencies; it acts as a state machine and persistence layer while the current session's model does all the actual writing, with no extra API calls.

Beyond the workflow, it exposes a set of style controls: register, tone, audience, language, perspective, and per-section or full-text word-count targets, all adjustable on the fly. Eight built-in templates cover WeChat articles, tech blog posts, weekly reports, novel chapters, Xiaohongshu notes, speeches, research reports, and user guides, each with a default outline and writing prompts that you can fork and customize. The assembled output is a clean Markdown file with front matter, ready to publish or commit.

If you use DeepSeek Harness for writing tasks that demand multi-section structure, iterative review, or cross-device continuity, this plugin moves your writing state out of session memory and into the filesystem, making resume-from-breakpoint feel as natural as opening a folder.