DeepSeek Harness Plugins
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
dsh-canvas-preview is a canvas preview plugin for DeepSeek Harness, enabling live preview of AI-generated HTML artifacts and one-click high-resolution export.
A one-click plugin that integrates the model-agnostic Codex web search MCP server into DeepSeek Harness, offering search, deep research, and text fetching tools.
dsh-agent-arena is a DSH coding arena for comparing 2–4 configured models in isolated Git worktrees, validating changes deterministically, reviewing diffs, and applying a winner.
A Git status visualizer for DeepSeek Harness (dsh) Web UI, featuring a session-header pill, a detail popover, and a full Git center without terminal.
The dsh-net-tools plugin enables DSH agents to securely fetch blocked websites and APIs via a local HTTP proxy in the sandbox, with tools for content fetching and proxy status checking.
worktree-mgr plugin provides task-isolated workspaces for dsh, automatically managing git worktrees to handle parallel tasks with independent development, commit, merge, and cleanup.
A DSH web plugin that adds a skills management page to the settings panel, allowing users to browse, edit, import, export, and delete skills.
Summarize a long DSH session with an LLM and auto-open a fresh session to continue seamlessly.
A DeepSeek Harness plugin for managing Python virtual environments within a workspace, featuring tools for discovery, creation, package installation, uninstallation, and removal with sandbox-safe subprocess execution.
A plugin for DSH that enhances the sidebar into a workspace folder list with task, file, and Git sub-tabs, plus a file preview panel.
A tool that recommends DeepSeek Harness plugins via natural language queries, providing quality scores, risk badges, and install commands, with full coverage and daily updates.
dsh-screenshot is a lightweight screenshot plugin for DSH, enabling one-click full-screen or area capture, Agent self-screenshot, path-only transfer, zero dependencies, and optional modlens integration for structured content reading.