dsh-plugin-skill-autoroute
`dsh-plugin-skill-autoroute` 是 DeepSeek Harness (DSH) 的插件，旨在解决技能被动触发不可靠的问题。它通过确定性路由扫描 Host 的实时技能目录，替代 LLM 进行决策，实现零额外调用的精准路由。插件提供 `brief`、`load` 和 `router` 三种模式，支持配置词典与评分策略。安装后需通过 `plugin_manager` 配置 `cordis.patch.yml` 文件以生效。该方案具备单次注入、实时扫描等特性，但存在仅支持全局技能层、依赖人工维护词典等限制。
Read More科研技能鉴赏：科研相关 Skill 的总入口与路由
科研技能鉴赏是 SkillHub 上的科研技能总入口与路由（@user_164f4c1f/keji-skill-showcase，v1.2.0）。当你不确定该用哪个科研技能时，可先启用它：精选 13 个技能覆盖文献检索与下载、知识库、CAJ 转 PDF、引用核查、学术诚信、组会、新颖性、答辩、方法论、技能锻造与 AI 易读化等。每个技能标注适用边界、能做什么、不能做什么、该不该装，并列出触发词。安装需按 skillhub.cn/install/skillhub.md 将技能装到 Cursor 等客户端的 skills 目录，无需 API Key。
Read Moredsh-skill-router: Agent-Driven Skill Discovery and On-Demand Loading
The DeepSeek Harness (DSH) skill routing plugin is designed to address token overhead caused by an excessive number of skills. By providing the `skill_search`, `skill_load`, and `skill_ref` tools, the agent can retrieve and load skills on demand, shifting the cost from a fixed charge each turn to a pay-only-when-used model. After installation, the agent can search for skills in the `.skill-src` directory, while the plugin includes a starter skill library to assist with usage.
Read Moredsh-skill-router: Rule-based skill routing before step execution
# Introduction When building agents with DeepSeek Harness (DSH), the timing of skill loading is usually left to the model to decide. Skills are exposed as a catalog to the model, visible at every step, forcing the model to re-evaluate whether to load them constantly. The result is unstable decision-making: skills may load too late, load the wrong one, or fail to load entirely. dsh-skill-router shifts this to a rule-driven approach: before each step, pure rules match the latest user message. If a match is found, the corresponding skill content is injected into the current step; otherwise, no intervention occurs. The entire process involves zero LLM calls and zero token costs. Below, we introduce its usage. # What is it? dsh-skill-router is a
Read Moreskill-bartender: Minimal Skill Allocation — Task-Skill Matching for DeepSeek Harness (Meta-skill)
# Introduction DSH's philosophy is that everything is a plugin. After loading many skills, the agent faces a constantly growing directory. While it can see the name and description of each skill, the common behavior is "over-dumping": loading too many skills, loading the wrong skills, or missing the workflow skill that was originally intended to string the whole task together. This presents an asymmetric cost: once a wrong skill is loaded, the skill content remains in the conversation history; whereas missing a load only results in the loss of a single tool call. Since missing is cheaper than being wrong, a reasonable strategy is to err on the side of loading less. skill-bartender is implemented according to this logic—it does not provide specific capabilities, but is only responsible for deciding "which skill this task should load, or whether to load it at all."
Read MoreDeepSeek Better Harness: A DSH Plugin Containing Three Evidence Swimlanes
`guyuefangyuanl/deepseek-better-harness` is an unofficial, zero-runtime-dependency DeepSeek Harness (DSH) plugin designed to integrate the three evidence lanes of Better Harness into a DSH profile. Maintained by an individual, it is MIT-licensed and has no official affiliation with DeepSeek or Huanfang. Its core functionality involves registering `session-evidence`, `project-readiness`, and `agent-assets` as three independent, routable skills, supporting the activation of specific lanes via `config.lanes` configuration, or the separate installation of individual lanes from `packages/`. The plugin does not open ports or start services.
Read More