DeepSeek Harness Plugins
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
dsh-factory enables durable task graphs and recurring agent work for DeepSeek Harness, transforming a single prompt into a dependency-aware production line with isolated checkouts, live sessions, and reviewable results.
A high-resolution vision enhancement plugin for DeepSeek Harness, designed for the deepseek-v4-flash-vision-exp model, which improves recognition of large images through automatic tiling.
The dsh-visualizer plugin streams and visualizes HTML documents in chat with live preview, download, and sharing capabilities, without server interaction.
A DSH plugin for parsing, querying, filtering, statistics, and converting CSV text with zero dependencies, pure functions, and an RFC 4180 state machine parser.
dsh-tool-diff is a text diff tool plugin for structured comparison and unified diff generation across text, JSON, CSV, and Markdown, featuring zero dependencies, pure functions, and read-only operation.
The dsh-tool-markdown plugin provides bidirectional conversion between HTML and Markdown, GFM table normalization, and table of contents generation, built with a hand-written parser, zero dependencies, and secure.
Provides a session persistence backend based on SQLite or PostgreSQL, with configurable database selection.
The dsh-tool-time plugin provides strict ISO time parsing, IANA timezone conversion, UTC calendar arithmetic, and fixed duration differences, implemented as zero-dependency pure functions.
Transform a paper or concept into an interactive research graph, tracing idea origins, evolution, and conflicts to deepen understanding of academic fields.
A K12 education skill that enables AI agents to access question retrieval, free random quizzes, practice tests, and intelligent test generation across subjects and grades for teachers, students, and parents.
reSanity is an evidence-search and logic-structuring skill for retail investment research, analyzing economic exposure chains, annotating evidence boundaries, and updating judgments with cognitive anchors.
deepseek-vl-support adds vision capabilities to AI models by automatically sending images to a vision service for text descriptions, enabling them to comprehend pictures like error screenshots and UI mockups.