dsh-multimodal-skill
v587d
Give text-only LLMs a pair of eyes. A DeepSeek Harness (DSH) native skill + zero-dependency Python CLI, adding image understanding and document parsing (OCR, tables, formulas, PDF → Markdown) to DeepSeek and other text-only models, using third-party multimodal APIs with free tiers first, direct connection on domestic networks, no proxy needed.