Community-maintained Agent Skills pack (MIT) for building DeepSeek Harness plugins — converges the official docs' scattered conventions into one enforceable standard, with minimal runnable examples and a trigger-eval regression set.
DSH integration
Ecosystem-related
Author-claimed
Safety audit
Unaudited
Last verified
2026-09-04
License
MIT
01What can it help you accomplish?
Develop DeepSeek Harness (DSH) plugins under one enforceable standard
Runnable DSH plugins and model tools, plus hello-plugin / greet-tool minimal examples
DSH plugin engineers and developers using Claude Code / Codex / DSH
Guard the skill description against trigger regressions
A trigger eval set (12 positives + 9 negatives) with an eval methodology
Skill maintainers and contributors
02How to install into DeepSeek Harness
Prerequisites
- A working DSH environment: Node.js, pnpm, and `dsh` (the skill itself needs no build or dependencies)
Installation steps
- 01
$ git clone https://github.com/zimodzh/dsh-plugin-dev-skills.git ~/.claude/skills/dsh-plugin-dev
- 02
Swap the target directory for your agent's skills path (DeepSeek Harness / Claude Code / Codex / VS Code Copilot, etc.)
- 03
Or download the release zip, unzip, and rename the folder to `dsh-plugin-dev`
Verify the integration
- Ask the agent "develop a DSH plugin / write a DSH tool" — the skill should trigger
- In DSH, load it directly with the `skill(dsh-plugin-dev)` tool to confirm
03DSH integration and capability boundaries
Loaded as an Agent Skills pack by DSH (and Claude Code / Codex and other compatible agents) to guide DSH plugin development under one shared standard.
DSH plugin dev standard library
An agent session that needs to build a DSH plugin→Unified plugin / service / event / tool / LLM-adapter conventions, a decision cheat-sheet and a pre-completion checklist
Minimal runnable examples
A DSH dev environment (Node.js / pnpm / dsh)→hello-plugin (bundle plugin) and greet-tool (model tool) that run and verify directly
Trigger eval set
Before and after editing SKILL.md description→Regression eval results (positive / negative pass rate)
04Who is it for? When not to use it?
Good for
- DSH plugin engineers and developers using Claude Code / Codex / DSH
- Skill maintainers and contributors
Not for
- Covers only in-repo, file-based DSH plugin development. Does NOT cover in-session dynamic plugins (cordis_define/cordis_run flows) or agent preset composition editing.
05Compatibility, maintenance and safety notes
- Covers only in-repo, file-based DSH plugin development. Does NOT cover in-session dynamic plugins (cordis_define/cordis_run flows) or agent preset composition editing.
- Community-maintained project with no affiliation with or endorsement from DeepSeek.
- Actually developing DSH plugins needs a working DSH environment: Node.js, pnpm and `dsh`. The skill itself needs no build or script dependencies.
MIT 许可证；社区维护，内容蒸馏自 2026-08 DeepSeek Harness 官方文档快照
06Frequently asked questions
What is dsh-plugin-dev and who is it for?
It is an Agent Skills pack (skill folder name `dsh-plugin-dev`, repo `dsh-plugin-dev-skills`) for developing DeepSeek Harness (DSH) plugins. It targets developers who build DSH plugins, services, tools, or LLM adapters inside DSH, Claude Code or Codex.
How do I install it?
Clone the repo into your agent's skills directory, e.g. `git clone https://github.com/zimodzh/dsh-plugin-dev-skills.git ~/.claude/skills/dsh-plugin-dev`, or download the release zip and rename the folder to `dsh-plugin-dev`. The skill itself needs no build or dependencies; developing DSH plugins additionally needs Node.js, pnpm and `dsh`.
How do I verify it loaded?
Ask the agent "develop a DSH plugin / write a DSH tool" — the skill should trigger. In DSH you can also load it directly with the `skill(dsh-plugin-dev)` tool.
What does it cover, and what is out of scope?
It covers in-repo, file-based DSH plugin development: plugin bundles, cordis.yml lines, patch overlays, tools, adapters, combo packs, profiles and in-repo workspace packages. It does NOT cover in-session dynamic plugins (cordis_define/cordis_run flows) or agent preset composition editing.
Is it officially endorsed by DeepSeek?
No. It is a community-maintained project with no affiliation with or endorsement from DeepSeek. Content is distilled from the official docs (2026-08 snapshot) and follows the "generated reference wins" principle.
07Related DSH workflows
distilly
by titanwings
Meta-skill turning source material into a versioned Person Profile skill with incremental merge, correction layer, rollback, per-host install.
dsh-market
by dsh-market
插件市场 (plugin marketplace): 打开 Settings → Plugin Market 浏览/搜索/一键安装社区插件；安装源限制为 curated awesome-dsh-plugin 注册表（其余拒绝），默认禁用构建脚本（pnpm>=10 需显式开启）。
ai_animation
by unclecheng-li
Generate standalone animated HTML diagrams, presentations, notes, protocol visualizations, and UI demos from prompts.
aegis
by ganyuanran
Make AI coding agents architecture-aware: baseline-first, evidence-verified, drift-checked, and safe across long tasks.
08Data and sources
一套遵循 [Agent Skills 规范](https://agentskills.io) 的技能，用于开发 [**DeepSeek Harness（DSH）**](https://github.com/deepseek-ai/deeps…
DSH 是一个插件化的 Agent Harness SDK：模型适配器、工具注册表、会话日志、甚至 agent loop 本身，全都是可以从配置里替换的 Cordis 插件。
本技能把[官方文档](https://deepseek-harness.github.io/deepseek-harness/guide/quickstart)里散落在教程、参考手册与生成目录中的约定，收敛成一套可执行的标准——任何加载了它…
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-09-04. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.