Preface¶

After creating a product, writing open-source code, and publishing articles, multi-platform promotion often gets stuck on two issues: the same message copied everywhere sounds like an advertisement, and manually rewriting for each platform requires restructuring and adjusting the tone. DeepSeek Harness (DSH) breaks capabilities into plugins for on-demand use; lhmd/dsh-promotion-toolkit specifically targets the workflow of “one source material, native output across multiple platforms.” Below, we’ll introduce what it does, how to install it, and how to use it.

What Is It¶

DSH Promotion Toolkit is a native DeepSeek Harness plugin released by maintainer lhmd, categorized as a workflow. In one sentence: it transforms product stories, open-source READMEs, articles, courses, events, or unorganized ideas into ready-to-publish promotional content for multiple platforms.

It doesn’t shorten a paragraph 17 times; instead, it first extracts core ideas, citable evidence, risk boundaries, and original links, then adapts the reading entry point for each platform:

Any idea / original text / README
            |
            v
Core idea + evidence + risk boundaries + original links
            |
            v
17 platform-native long-form posts, short posts, video scripts, and CTAs

License: MIT. GitHub repository: https://github.com/lhmd/dsh-promotion-toolkit

Core Features¶

Platform-Differentiated Output¶

The same fact is presented with different structures on different platforms, not just summarized uniformly. The README lists the coverage as follows:

Total of 17 platform identifiers: xiaohongshu, zhihu, wechat, wechat_moments, weibo, bilibili, douyin, kuaishou, x, threads, linkedin, reddit, tiktok, youtube, medium, facebook, instagram.

Structured JSON Output¶

Each call returns a structured result containing core ideas, evidence, shareability signals, title angles, drafts for each platform, image prompts, tags, and risk flags. The model only polishes and adapts for platforms, and should not add users, customers, metrics, dates, prices, results, or links that aren’t in the original text.

{
  "coreIdea": "Core idea extracted from the original text",
  "evidence": ["Evidence to preserve"],
  "shareability": {"score": 0, "signals": []},
  "titles": ["Title angle one", "Title angle two", "Title angle three"],
  "platformDrafts": {
    "xiaohongshu": "Saveable note",
    "zhihu": "Markdown explainer",
    "wechat_moments": "Friend-style share"
  },
  "riskFlags": ["Verify before publishing"]
}

DSH Integration and CLI¶

The repository uses the DSH profile-bundle structure: cordis.patch.yml mounts the package, src/index.js exports apply(ctx) and registers the viral_kit tool, src/skill.js provides runtime Skills, and skills/viral-kit/SKILL.md is a human-readable mirror. It also provides a standalone CLI: scripts/viral-kit.mjs (npm package name @lhmd/dsh-promotion-toolkit, current version 0.1.0).

Installation and Enabling¶

First, install the DSH CLI, then add the plugin directory to your profile:

npm install -g @deepseek-ai/dsh
dsh plugin --profile web add /absolute/path/to/dsh-promotion-toolkit
dsh web

Replace /absolute/path/to/dsh-promotion-toolkit with the actual absolute path after cloning or downloading on your local machine. The plugin runs with the current dsh process permissions; you should review the source code and MIT license before installation.

Local development and checks:

npm test
npm run check:release
npm run demo

Typical Usage¶

Command-Line Single-Platform Generation¶

The following command generates a promotional draft for a specified platform from a text input and outputs it as JSON:

node scripts/viral-kit.mjs \
  --text "Paste your product introduction, article, course, event, or any idea" \
  --language auto \
  --platform xiaohongshu \
  --source-url https://github.com/lhmd/dsh-promotion-toolkit \
  --json

--language supports zh, en, auto. Chinese platforms default to Chinese; global platforms default to English; explicit input takes precedence. --source-url is for passing in repository, product page, article page, registration page, or other original links—the plugin places it in the corresponding platform’s CTA without fabricating links.

Reference Examples and Demo Page¶

The repository includes full examples generated using deepseek-v4-pro: 4 long-form topics, in both Chinese and English, with 17 platform versions each, preserving original links in each draft.

• examples/live-demos.json: Complete structured JSON
• examples/live-demos.md: Each platform displayed in code blocks for easy copying
• Online bilingual demo page: https://lhmd.top/dsh-promotion-toolkit/ (switchable between Chinese/English, 4 cases, and 17 platforms)
• docs/: Demo page source files

Use Cases and Notes¶

Suitable for developers, indie developers, and content creators who need to distribute the same source material across multiple social or content platforms: open-source project README promotion, product launches, course or event promotion, long-form articles adapted for different platforms, etc.

When using, note:

1. Verify attribution, numbers, time, and absolute statements before publishing; riskFlags will highlight content that needs manual review.
2. The plugin won’t prove dissemination results for you, nor should it fabricate users, revenue, dates, or data.
3. Platform rules and formats are still evolving, so drafts should undergo a final check according to platform guidelines.

Conclusion¶

DSH Promotion Toolkit transforms “multi-platform promotion” from copying the same text to first extracting ideas and evidence, then adapting the reading path for each platform. If you’re using DSH for workflows, you can integrate this chain into your profile and start from one original text using the CLI or viral_kit tool.

• SkillHub catalog page: https://www.skillhub.cn/plugins/lhmd/dsh-promotion-toolkit
• GitHub repository: https://github.com/lhmd/dsh-promotion-toolkit