DeepSeek Harness Plugins
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
dsh-ultramath is a multi-agent plugin for mathematical modeling competitions, using DeepSeek Harness presets to automate the solving process from problem reading to paper writing.
dsh-crew transforms a DeepSeek Harness session into a product manager that orchestrates a crew of role agents for collaborative tasks, supporting both inquiry and team workflows.
Integrate wallpapers downloaded in Wallpaper Engine as the background for the DeepSeek Harness web interface.
A DSH plugin for self-improving AI agents that enables continual learning through persistent memory, periodic review, cross-session knowledge sharing, and automatic rollback on failure.
A DSH plugin that ports Remotion for React-based video creation, supporting animations, audio, subtitles, and more.
dsh-md-quiz is a quiz-based reading plugin for DeepSeek Harness that transforms Markdown files into interactive learning sessions with 5-stage challenges, diagnostic tests, and progress tracking, accessible via a Web UI.
dsh-remote-ssh is a DSH plugin similar to VSCode Remote-SSH, enabling SSH connections to remote supercomputers or servers for direct manipulation of remote files and terminals within the built-in tabs.
Real-time display of DeepSeek account balance in the sidebar, supporting multiple currencies, with settings page for balance trends and usage statistics.
A collaborative kanban board that allows you and your AI agent to plan and manage work together within DeepSeek Harness, without leaving the environment.
dsh-a2a provides a realtime Agent2Agent mesh for DeepSeek Harness, enabling message sending, history tracking, and project management for agent collaboration on a trusted private network.
Multi-host coding CLI that lands one declarative adapt table natively across 13 hosts with P0 gates, Harness process commands, and unified coding-standards injection.
Controls how the model calls tools rather than its output: injects behavioral soft constraints, converges parallelism on failure, and snapshots files before writes for safe rollback.