DeepSeek Harness Plugins · Workflow
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
AI-driven UI automation plugin for DeepSeek Harness powered by Midscene, supporting Android and Chrome via natural language.
Connects independent agent windows into a flat team, enabling structured task dispatch and real-time progress tracking.
A DeepSeek Harness workflow plugin that records checkpoint markers to fold exploration noise into reports using the rewind plugin.
Provides a structured Remotion workflow and five workspace-confined tools for diagnosing projects, rendering videos, and verifying output metadata.
Integrates ChunSun's AI project delivery platform into DeepSeek Harness, enabling autonomous project delivery via requirements and real-time progress monitoring.
Notifies users during task completion, errors, running, and waiting states, supporting system notifications and desktop popups with click-to-jump session functionality.
Supports viewing and interacting with OBJ/STL/STEP models in the DSH Web interface and automatically translates them into standard JSON readable by LLMs.
A self-evolving plugin for DeepSeek Harness that observes agent signals, proposes improvements via LLM, and automatically applies validated changes with rollback.
dsh-harness-ally enables Harness Alliance mode, allowing DeepSeek Harness, Claude Code, Codex, and Kimi Code to share DSH session lifecycle with DSH's native model selector.
A DeepSeek Harness plugin for isolated, read-only Git diff reviews using Pi Agent, returning structured findings with severity and usage metrics.
An AI course-authoring workspace plugin for DeepSeek Harness with a three-pane interface, nine-phase gated workflow, course/chapter/lorebook management, knowledge graph, TXT/Word export, and sharing capabilities.
A persistent DAG workflow kernel embeddable into any Agent Host, orchestrating agents, tools, skills, and MCP into a saveable, verifiable, and replayable workflow template JSON.