Preface¶
The DSH plugin ecosystem follows an “everything is a plugin” approach. As the number of plugins grows, the real problem isn’t “whether they exist” but “how to find the right one.”
dsh-plugin-advisor brings this problem directly into DSH conversations: describe your needs in natural language, and it returns plugin recommendations with quality scores, risk badges, and installation commands. It targets both DSH users and agent developers, requiring no additional LLM API calls for recommendation ranking.
The DSH community directory is an independent site with no official affiliation with DeepSeek / High-Flyer, and it should not be interpreted as an official app store.
What This Is¶
dsh-plugin-advisor is maintained by ChengxiuCDP and licensed under MIT.
One-line positioning: quality-aware DeepSeek Harness plugin discovery.
It addresses several issues when searching for dsh-plugin within DSH: candidate matching, quality and risk indication, and installation command generation.
Core Features¶
Natural Language Queries¶
State your needs directly in DSH conversations, and the agent will invoke the find_dsh_plugins tool to search for plugins.
Returned Results¶
The returned content includes:
- Top matches
- Matching rationale
- Risk warnings
- Installation commands
Ranking and Recommendation Rationale¶
Ranking is a pure rule engine with no LLM API calls.
The recommendation rationale is generated by the agent during its own reasoning turn.
The built-in index comes from the complete dsh-plugin topic repository, with 4700+ entries, used as a daily index snapshot.
Quality Score Dimensions¶
Quality score dimensions include:
- Installability declaration
- Curated inclusion
- Stars
- Activity level
- License
Matching and Freshness¶
Built-in feature clusters and feature alias matching are used.
When the gh CLI is available locally, each query is augmented with a real-time GitHub search.
Real-time GitHub search is only applied when the gh CLI is present on the local machine.
Public Site¶
A public site is available for online search, allowing you to experience the same dataset and ranking engine directly in your browser.
Installation and Activation¶
Environment requirements:
Node.js >= 20.0.0
Installation command:
dsh plugin --profile web add -w github:ChengxiuCDP/dsh-plugin-advisor
Restart dsh web after installation for it to take effect.
If pnpm does not report a workspace root prompt, you can omit the -w flag.
Typical Usage¶
Enter the following in a conversation:
Find me a plugin that lets text-only models read images
I want to check session progress on my phone
Is there a Claude Code-style terminal interface?
The agent will invoke find_dsh_plugins and return top matches, matching rationale, risk warnings, and installation commands.
Risk warnings may include NOASSERTION, missing licenses, or discontinued updates.
Suitable Scenarios and Notes¶
Suitable for:
- Users who need to find plugins within DSH conversations
- Developers who need to quickly extend agent capabilities
- Those who want to review quality scores and risk warnings before deciding whether to install
Notes:
- Installing a plugin means running third-party code on your machine with your permissions.
- Plugins are executed by the current dsh process with local permissions; review the source code and license before installation.
- This tool only provides recommendations and warnings—it never auto-installs anything.
- Please review the source code yourself before installation.
- Recommendations do not constitute security audits.
Related Links¶
Directory page:
https://www.skillhub.cn/plugins/ChengxiuCDP/dsh-plugin-advisor
GitHub:
