xby-cookie-reward: Gamified Cookie Rewards for LLMs via Self-Reflection
`xby-cookie-reward` is a gamification plugin for DeepSeek Harness (DSH), designed to provide instant incentives for LLMs through a self-reflection mechanism. It allows LLMs to reward themselves with Cookies by evaluating their own outputs, and offers features such as setting API keys, checking balances, and resetting counters. Users must register on the Xiaobenyang website to obtain an API key. It can be installed with the command `dsh plugin --profile web add github:xby_skill/xby-cookie-reward`. Please note that adding Cookies to the reward jar is a human-only operation, and LLMs are not allowed to invoke this function.
Read Morexby-chucknorris: DeepSeek Harness (DSH) LLM Enhanced Service Plugin
`xby-chucknorris` is a model inference plugin for DeepSeek Harness (DSH), designed to provide enhanced prompts and system instructions to LLMs via MCP server mode, particularly suitable for security research and evaluation. The plugin offers three core functions: setting an API key, retrieving optimized prompts, and obtaining advanced system instructions to improve reasoning capabilities. Users must obtain the API key from the official website, and installation is recommended via terminal commands from GitHub. Additionally, the plugin is open-sourced under the MIT license, and developers should be mindful of permissions and security specifications when using it.
Read Moredsh-plugin-subs: DeepSeek Harness Subscription Integration Plugin
`dsh-plugin-subs` is a DeepSeek Harness (DSH) plugin designed to connect directly to consumer subscriptions such as Claude and ChatGPT Plus through an OAuth flow, eliminating the need to manually configure API keys. It supports model parsing, streaming output, and tool calling, and is easy to install. Users need only connect and authorize the provider in Settings to use their existing subscriptions for model inference in DSH. The plugin is suitable for users who want to simplify API management, but please be aware of provider-specific limitations and configuration requirements for Chinese platforms.
Read Moredsh-clamp: Numerical Clamping
DeepSeek Harness (DSH) `dsh-clamp` is a modular plugin focused on numerical processing, designed to constrain model outputs through value clamping and prevent anomalous numerical values from affecting downstream logic. It is maintained by uckkk, implemented in pure Node, has no network dependencies, and runs efficiently. Installation requires the specified version requirements of `@deepseek-ai/cordis` and `@deepseek-ai/dsh-tools` to be met. After installation, users can directly invoke the tools it provides within a session. It is recommended to review the source code and the MIT license before use, and to ensure the environment satisfies the dependencies.
Read Moredsh-llm-longcat: A Long-Context and Thinking Mode Adapter for DeepSeek Harness
dsh-llm-longcat is a DeepSeek Harness (DSH) plugin designed to integrate the LongCat-2.0 model. It primarily resolves issues in the Harness ecosystem related to long-context (1M token) processing, structured chain-of-thought display, and native tool calling. The plugin supports streaming, multi-turn conversation reasoning review, and API key management via environment variables. Installation requires running a local script, and configuration should note that image input, stop sequences, and binary limits on reasoning strength are not supported.
Read Moredsh-llm-ollama: DeepSeek Harness Ollama Provider Plugin
`dsh-llm-ollama` is a community Ollama provider plugin for DeepSeek Harness (DSH), used to integrate local or remote Ollama services within the DSH web profile. The plugin implements conversation via Ollama's native `POST /api/chat` protocol, supporting streaming responses, tool calling, multimodal image attachments, and reasoning process display. Core functions include a full-page settings interface for managing parameters such as the API address and temperature; one-click discovery and addition of models using `GET /api/tags`; mapping the context window and maximum output tokens to Ollama's `options.num_ctx` and `num_predict`; and auto-detection or manual control of the reasoning level via `thinkingCapable`. Furthermore,
Read More