Introduction

The core philosophy of DeepSeek Harness (DSH) is “everything is a plugin.” As the ecosystem grows, there are already 2600+ repositories under the dsh-plugin topic. Facing a massive plugin list, developers need an efficient way to discover, filter, and obtain plugins. awesome-dsh, as a standalone directory and search tool, addresses the challenge of finding plugins. It not only provides a static directory, but also includes a built-in Skill for automatically translating descriptions with Workers AI and on-demand search, helping developers quickly locate target plugins.

Core Features

  • Full Directory Crawl: Automatically scrape the full plugin directory under the GitHub topic dsh-plugin, supporting 2600+ plugins.
  • Automatic Translation: Use Workers AI to automatically translate English plugin descriptions into Chinese, with a permanent caching strategy that translates only new or changed entries.
  • On-Demand Search: Provide mixed Chinese-English term search; categorization is based only on the original text, while search matches both the original and translated text.
  • API Support: Provide endpoints such as /api/search, /api/plugins.json, /api/facets.json, and /api/status.json, supporting full snapshots, category tags, crawl progress queries, and manual triggers.

Installation and Enablement

This plugin provides two enablement methods; choose one based on your environment.

Install into DSH:
Run the following command to register the plugin in DSH:

dsh plugin --profile <profile> add github:stakeswky/awesome-dsh

Install into Claude Code:
Copy SKILL.md to the local skills directory:

mkdir -p ~/.claude/skills/dsh-plugin-finder && cp SKILL.md ~/.claude/skills/dsh-plugin-finder/

Typical Usage

After installation, you can trigger search directly through conversation or make programmatic calls via the API.

Conversational Search:
Directly state your request to Claude Code or DSH, for example:

“Install a plugin for DSH that can view images”
“Does DSH have any memory-related plugins?”
The system will search the local directory and return matching plugins and installation commands.

API Search:
Use the /api/search endpoint. Parameters include q (query), kind (type), fn (repository name, repeatable), and limit (number of results; default 8, maximum 25).

curl "https://dsh.jamintextiles.com/api/search?q=vision&limit=5"

Precautions

  • GitHub Rate Limits: Anonymous scraping of the GitHub Search API is limited to 10 requests per minute. For stable full crawling, configure GITHUB_TOKEN.
  • Resource Costs: Workers AI and KV storage both require a paid Workers plan; full crawling exceeds the free plan’s CPU and subrequest quota.
  • Package Activation Rules: Packages that do not declare dsh.bundle are installed only as regular dependencies and do not activate the plugin layer.
  • Build Restrictions: The first add of a git-source plugin is blocked by pnpm’s allowBuilds; manual handling is required.
  • Translation Strategy: Translation uses permanent caching and translates only new entries or entries whose original text has changed. If an individual translation fails, it is skipped and retried without affecting the overall process.
  • Search Logic: Category tags match only the English original text, while search matches both Chinese and English content.

Appendix

  • Online Directory: https://dsh.jamintextiles.com
  • Project Home: https://github.com/stakeswky/awesome-dsh