dsh-oil-creator

Anyone who makes videos knows the feeling: topic notes live in one app, scripts in another, while recording projects, subtitles, covers, and publishing status are scattered everywhere. AI can help draft scripts or generate covers, but without context it can hardly collaborate on a piece of content from start to finish. dsh-oil-creator tries to solve this by turning “one video” into a visible, coherent unit of work inside DeepSeek Harness. The core idea is simple: one video is one local folder. From topic and script to screen recording project, subtitles, covers, and publishing status, everything lives in the same directory. The plugin only stores workbench state—project bindings, publishing markers, playback data, generation tasks—in overlay.json. It never copies your content or locks it into a proprietary database. Any editor and any AI file tool can read your files directly.

The workbench organizes a video’s lifecycle into five stages: topic & script, recording & editing, subtitles & covers, publishing, and data collection. AI handles creating well-structured directories, reading and writing topic.md and script.md, binding Screen Studio projects, launching subtitle workflows, generating covers in three aspect ratios, preparing drafts for multiple platforms, and syncing views, likes, comments, and links via Ego Browser. Humans keep the final say on topic direction, timeline editing, voice, and the “publish” button on every platform. The plugin follows a “preview first, confirm second, execute third” principle—folder cleanup and batch renaming never touch files by default.

Key capabilities include a local library that scans directories named date_readable-title and shows stage, final cut, subtitles, covers, article, and publishing status; conversation context through @current details, content search, or /current content; AI bootstrap configuration that discovers standard install paths and suggests clear install steps when abilities are missing; persistent script rules that save tone, structure, taboos, and target audience for future script.md drafts; and long-running task tracking that returns immediately after subtitle, cover, or burn-in tasks start while the plugin keeps watching file outputs. Subtitles, covers, Screen Studio editing, multi-platform publishing, and data collection are all optional—install and configure them as needed. The core library and script management work without any external service.

This plugin is for DeepSeek Harness users and teams who want to keep their content creation workflow local, file-based, and transparent. If you want AI to push work forward and handle repetitive chores, but don’t want your scripts and final videos locked inside a proprietary database; if you need to open your files in any familiar editor, keep the final decision before publishing, and later sync public performance data for review—dsh-oil-creator offers a middle ground between fully automated pipelines and fully manual organization. It’s made for people who believe their content belongs to them, and AI is just a collaborator.