DeepSeek Harness Plugins
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
A line-anchored edit tool using variable-length Base62 content anchors — no re-typing old code, verified edit ranges, fewer tokens, and atomic batch edits with session-stable anchor chaining.
Securely expose the DSH Web UI to your LAN so a phone scan controls your desktop agent, with a built-in fork_session tool. Zero dependencies, zero core changes.
A floating DualSense controller overlay for the DeepSeek Harness web GUI that highlights pressed buttons and fires vibe-coding instructions directly from the gamepad.
A DeepSeek Harness web plugin that expands all Think lines in one toggle, persists across sessions, and respects manual collapse.
Records token usage per model call, polls DeepSeek balance on a timer, and displays a draggable live status card in DSH Web.
Cross-platform model search with real-time keyword filtering and one-click session switching for DSH multi-provider setups.
Guides agents in authoring reusable skills and isolates workspace-level skills so project-private ones never leak across workspaces while overriding globals by name.
A graphical skill manager for DeepSeek Harness — search, install, enable, disable, and remove skills from the Web UI with multi-channel support and zero file editing.
Lets the agent list containers, read logs, inspect images, and manage Compose projects via the machine's Docker CLI.
Serves the Monaco Editor distribution over a host HTTP route so Harness plugins can embed a real code editor in-page, with no CDN or third-party origin.
Intercepts delete commands (rm, Remove-Item, etc.) and turns every deletion into a recoverable trash move, with one-click restore and a single confirmed purge path.
Detects capability gaps, retrieves and audits DSH plugin candidates, and performs controlled version-locked installation with local, auditable records.