DeepSeek Harness Skills — dsh Plugins on the Cordis Architecture
The DeepSeek Harness (dsh) is an open-source agent harness with a Cordis-based plugin architecture. This page indexes the skills/plugins that ship with dsh and cross-links the existing /tools/deepseek-harness guides for install and provider setup.

DeepSeek Harness Skills
The DeepSeek Harness (dsh) is DeepSeek AI's open-source agent harness — a Cordis-based plugin system that ships its own set of skills/plugins. Unlike the cross-vendor skills in the rest of this collection, dsh skills are tightly coupled to the dsh runtime and won't run unmodified in Claude Code or OpenCode.
The site's existing DeepSeek Harness coverage carries the install, provider-setup, and TUI configuration guides. This page is the skills index that cross-links to that coverage so dsh plugin content lives next to the rest of the collection.
When to use
- You're running dsh as your agent runtime and want the canonical plugin index
- You maintain a dsh plugin and want it discoverable from the broader skills collection
- You want a single reference for "what skills exist in dsh and where do they live"
When not to use
- You're not on dsh — the dsh skills are runtime-specific. For cross-vendor skills, see the rest of this collection.
- You want Anthropic-style SKILL.md portability — dsh plugins are Cordis plugins, not agentskills.io skills (yet)
How it works
dsh plugins live in the dsh install and are activated by the harness on relevant triggers. The plugin manifest and per-plugin code is in the dsh source tree (the npm package is @deepseek-ai/dsh; verify the source repo from the package metadata before installing). The site's existing guides cover how to install dsh and configure providers; the plugin list itself is what dsh ships — check dsh plugin list after install.
Install
Once dsh is installed (see the DeepSeek Harness getting-started guide):
dsh plugin install <plugin-name>
For TUI and provider configuration, see the existing DeepSeek Harness TUI guide and Providers & Keys guide.
Full plugin source
Plugin manifests and code live in the installed @deepseek-ai/dsh package and the upstream source repo (verify the canonical repo URL from the package metadata before installing). Embedded copy omitted at launch — dsh plugins are runtime-coupled and the canonical source is the right reference. Verify each plugin's license before redistribution; dsh core is open-source but individual plugin licenses vary.
Verify it yourself
Quick static check — dsh plugins may legitimately make network calls (the harness is a coding agent); confirm the endpoints are DeepSeek-controlled and not third-party. Bounded allowed-tools is a positive signal.
Two-model review — paste each plugin's manifest + code into two different models with the Vetting Guide prompt. dsh plugins are the highest-capability skills in the collection (full shell, file, network in a coding-agent runtime) and deserve the most scrutiny.
Review transparency
| Field | Value |
|---|---|
| Last reviewed | 2026-08-28 (launch baseline; cross-link to existing coverage) |
| Reviewers | site baseline (static scan + cross-link verification only) |
| Source repo | npm: @deepseek-ai/dsh (verify upstream repo from package metadata) |
| License | verify per-plugin (dsh core open-source; plugins vary) |
| Re-vet trigger | any plugin change; >30 days since review |
Related
Related Articles & Guides
#skillsSkills Collection — Curated Agent Skills for Every Tool
A curated, vetted catalog of agent skills (SKILL.md) that work across Claude Code, OpenCode, Gemini CLI, Codex, Cursor, and 40+ other agents. Each skill includes install tabs, full source, and instructions for verifying it yourself.
#skillsHow to Vet an Agent Skill With Models
A repeatable methodology for vetting agent skills before installing. Static red-flag scan plus a copyable two-model review prompt that flags prompt injection, exfiltration, and over-broad permissions. Built on the Aug 2026 security research wave.
#skillsAnthropic Official Skills — Reference Implementations of the Standard
Anthropic's published skill examples — artifacts-builder, pdf, docx, and others. The canonical reference for SKILL.md anatomy, frontmatter conventions, and progressive-disclosure patterns.