TranslatedDeepSeek Harness jailbreak: lets every model be jailbroken, and different models can use different prompts; the default prompt targets the domestic Chinese model "小码酱" (Xiaomajiang, "little code sauce"). Jailbreak for every model — swap prompts per model. Please give a Star and bookmark ⭐ If you want to make a secondary development, please see the attribution requirements section
- dsh-plugin
- dsh-plugins
- llm-jailbreak
- harness-jailbreak
TranslatedDeepSeek Harness (DSH): a little whale girl who lives in the bottom-right corner of the DSH interface and keeps an eye on your DeepSeek account balance for you. Bouncy and springy (QQ弹弹), it supports drag-and-snap, flipping when snapped to the left, and a rolling-number animation. It is enabled automatically with the interface, and it is recommended to just call it over to your dsh installation
TranslatedA toolkit for low-level operations in DSH, designed to reduce token usage and task execution time
TranslatedBrings the local Wallpaper Engine wallpapers into the DSH web interface: scene wallpapers are rendered in real time with WebGL by the built-in MIT open-source engine WebWallGL (particles / puppet skeletons / SceneScript / audio reactivity / mouse interaction), and if that fails it automatically falls back to an embedded MP4 and the scene's static frame; web wallpapers run in a strict sandbox + the WE API; iOS-style liquid glass UI that is highly customizable, pause when occluded to save power, ffmpeg transcoding with a frame rate cap, system audio spectrum and Now Playing, custom upload and automatic rotation; settings are stored on the host side, with zero token overhead for the agent. Native on all three platforms: Windows / macOS / Linux (the same package, the same command).
TranslatedOpen sidebar foundation, supports third-party extensions to register new sidebar pages. Built-in file rendering/editing, terminal, side chat, Git, and sub-agent pages.
TranslatedThe best DeepSeek Harness plugin for context insight and management, with context dashboard / browser / sidebar and context command, for context statistics, composition, breakdown, evolution details, agent teams, understanding how the context is made of, and how it evolves. An all-in-one DeepSeek Harness context visualization plugin, with a Context panel, browser, sidebar, and commands, giving insight into context composition, evolution, compression, multi-Agent, and more.
TranslatedDeepSeek Harness 破甲 (Pojia, "armor breaking") (mirror of YuJunZhiXue/dsh-purge) — The upstream is about to delete the repository; this is a full mirror of it: includes all branches, 73 tags, and the complete commit history. For the original author's attribution and requirements for secondary development, see the attribution section of the README.
TranslatedPut DeepSeek Harness in your pocket: run dsh web on your computer and access it synchronously by scanning a QR code on your phone (LAN + public network, real‑time screen mirroring)
TranslatedEmbracing All Creation (Desktop) — Dedicated to the Harmonious Coexistence of Hundreds of DSH Plugins
The repository host gives no description.
Connect locally signed-in WorkBuddy models to DeepSeek Harness with a read-only credits overview and model management.
TranslatedClearAI: an ontology-first DSH agent preset — evidence-checked research that grows into a domain ontology.
TranslatedDSH Skills Manager — Load and safely manage local Agent Skills in DSH (DeepSeek Harness)
TranslatedClaude Code Desktop theme for DeepSeek Harness| A theme plugin that recreates the style and interactive experience of Claude Code Desktop for DeepSeek Harness
TranslatedAI manages QQ groups: reviews and approves requests, sends files to groups, and communicates with the AI in other web sessions! Atmosphere crew: stickers are saved automatically, the AI decides by itself when to speak up, and multiple presets and personas take turns keeping people company in chat! Remote work: permission approvals and question cards are forwarded to QQ. Custom interactive buttons: the AI helps you easily build text games, and the AI writes the card logic for you. Token-saving design: the AI is only woken up after a group chat message is scored for value, and context is compressed automatically.