Preface

The core philosophy of DeepSeek Harness (DSH) is “everything is a plugin.” When developing agent applications, transforming a vague user requirement into a structured, directly deliverable document often requires precise orchestration. The following section introduces the dsh-resume-expert plugin, which solves problems of structural confusion and cumbersome layout in resume generation through a conversational workflow.

What It Is

dsh-resume-expert is a guided resume-generation plugin maintained by wmw343. Through a four-stage conversational workflow, it transforms vague requirements into a single-page A4 PDF that can be submitted directly.

Core Features

Four-Stage Conversational Workflow

The plugin uses 6 tools to divide the process into four stages:

  1. Initial Diagnosis: Input a vague requirement, then generate a first text draft, a completeness assessment, and a list of missing items.
  2. Iterative Completion: Absorb supplementary user information across multiple rounds (name, education, internships, projects, skills), and proactively ask for key information.
  3. Final Draft: Integrate everything into a final version that can be submitted directly, removing placeholders.
  4. Export: Generate an A4 PDF (falling back to printable HTML if Chrome is not available on the local machine).

The 6 involved tools include resume_intake, resume_refine, resume_compose, resume_export, resume_set_photo, and resume_tailor.

Output and Customization

  • Direct A4 PDF Output: Generates a real-layout single-page PDF that can be submitted directly.
  • ID Photo Addition: Supports uploading a local image to the top-right corner of the resume.
  • JD Tailoring: Rewrites resume content according to a target job description (JD).

Quality Control

The plugin has built-in resume quality rules, including quantified phrasing, bolding key numbers, skill grouping, pinning key metrics (≤3 items), primary/secondary city ordering, phone-number 3-4-4 formatting, and more.

Acceptance Criteria

The plugin includes 471 automated acceptance checks. The exported PDF is validated through physical parsing by pypdf, checking page count, paper size, structure, and absence of placeholder residue.

Installation and Enablement

Prerequisites: DeepSeek Harness ≥ 0.1.5, Node.js ≥ 22.

dsh plugin --profile web add github:wmw343/dsh-resume-expert#v1.0.1

Configure environment variables so the host environment can call the DeepSeek API:

# 写入 DSH 主目录的 .env
Add-Content "$env:USERPROFILE\.dsh\.env" "DEEPSEEK_API_KEY=sk-你的Key"

Typical Usage

After starting dsh web, enter the following in the conversation:

“Help me make a resume. I am a new graduate from [major] and I know [skill A] and [skill B].”

The plugin will automatically enter the initial diagnosis phase, then guide you through completing the information, and finally output an A4 PDF.

Data and Security

  • Key Management: The plugin does not hold keys; model credentials are provided by the host environment.
  • Data Storage: Conversation state is persisted by default to ~/.dsh/resume-expert-store, without database storage or external transmission.
  • Cross-Process Persistence: Supports continuing to modify the draft after restarting dsh web.

Development and Self-Testing

The plugin includes an offline self-testing feature, requiring no network or DSH startup:

npm install && npm run build   # 编译
npm test                       # 运行 21 项存储不变量测试

Use Cases and Cautions

Suitable for users who need to quickly generate a structured resume. The plugin runs with the current DSH process permissions; it is recommended to review the source code and license before use.

Conclusion

dsh-resume-expert provides a complete closed loop from conversation to PDF, making it suitable for integration into workflows.