dsh-vision-guard: Transparent Image Protection and Visual Analysis Tool for DeepSeek Harness

`dsh-vision-guard` is a transparent image protection and visual analysis tool designed for DeepSeek Harness. This plugin addresses the deadlock issue in text-only models caused by image processing by converting images into text before they are written to conversation logs, and it can also recover sessions that have already deadlocked. Its core features include automatic image-to-text conversion, support for document and video analysis, native image passthrough, and zero dependencies. To use it, a vision model must be configured; document processing requires installing the corresponding system tools, and images must be within 5 MB.