DeepSeek Harness Plugins · Model Inference
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
An offline LVGL 8.4 knowledge pack for DeepSeek Harness teaching the agent to write correct v8 code and prevent v9 API mixing.
Bridges DeepSeek with Google Gemini multimodal capabilities for vision, generation, and editing with self-check loops, requiring no model switching.
dsh-prototype is a DeepSeek Harness model inference plugin for future development.
dsh-cad-modeling is a DeepSeek Harness plugin for CAD model inference.
DeepSeek Harness bundle adding an Auto Agent Preset. It classifies prompts with V4 Flash and routes them to standard, code, minimal, or cordis presets, automatically selecting tools and skills.
A DeepSeek Harness plugin offering six econometrics tools for the complete research workflow.
Automatically routes subagent tasks between Flash and Pro for cost savings.
Hybrid auto model router for DeepSeek Harness, automatically switching between L0-L3 model levels based on task complexity with budget control.
Enables parallel candidate generation and automatic selection to boost task success rates.
A beginner-friendly DeepSeek Harness plugin offering one-click installation, free model configuration (Zhipu GLM & OpenRouter), and cost monitoring for zero-barrier usage.
DeepSeek Harness plugin for multi-model governance, offering access control, credits, and smart routing strategies.
Turn any Codex/Claude skill repository into a native DSH skills marketplace managed by plugins.