Introduction

The core principle of DeepSeek Harness (DSH) is “everything is a plugin.” When building agent applications, user research is one of the core workflows for product managers and R&D teams. When dealing with large volumes of transcribed interview recordings, open-ended survey responses, or customer support feedback, letting AI process them directly can easily produce “hallucinations,” or yield conclusions that lack methodological support and are hard to apply.

dsh-plugin-user-research is a vertical-domain user research synthesis plugin. It provides a set of validated methodology tools and skills, helping developers or researchers quickly transform scattered raw materials into structured user personas, pain point lists, and opportunity maps.

Plugin Overview

This plugin is maintained by the developer nsdmgt and classified as a “workflow” plugin. Its core value lies in transforming unstructured research notes into actionable product decision-making references through standardized frameworks.

Core Capabilities

The plugin provides the following capabilities, helping users extract useful information from raw materials:

  • Topic Classification Criteria: It includes a built-in topic classification system, covering efficiency, ease of use, cost, feature gaps, stability, and service support. The system can automatically categorize scattered user quotes under these dimensions.
  • Anti-Hallucination Constraints: When extracting pain points, the tools strictly limit output to direct quotes from users, without expansion or speculation, ensuring conclusions are traceable.
  • Opportunity Scoring Criteria: It provides a set of opportunity scoring logic, sorted by signal strength. Scoring is based on mention frequency and whether users explicitly expressed pain, helping identify higher-priority directions.
  • Research Skills and Templates: It includes 4 reusable research skills (interview analysis, persona construction, opportunity mapping, and the full synthesis workflow), as well as 2 templates (interview outline and synthesis canvas).
  • Tool Invocation: It provides a tool interface named research_synthesize, supporting direct invocation within an Agent session.

Installation and Enabling

Installing this plugin requires the DeepSeek Harness CLI tool.

dsh plugin --profile web add dsh-plugin-user-research

After installation, you must restart the web service for the plugin to take effect.

Usage

Using this plugin usually involves three steps:

  1. Prepare Materials
    Organize transcribed interview recordings, exported open-ended survey responses, customer support chat logs, or user feedback. Supported file formats include .txt, .md, and .csv.

  2. Invoke the Tool
    In a DSH session, directly describe your requirements to the Agent. The Agent will parse the instruction and invoke the underlying research_synthesize tool.

    For example, enter the following:

    Run a user research synthesis on notes/interview-01.md and pick the top 3 opportunity points.

    The Agent will execute a call similar to:
    research_synthesize(filePath="notes/interview-01.md", maxOpportunities=3)

  3. View Results
    You will receive structured output, including:

    • personas: Inferred user personas (roles, demand signals, topics).
    • painPoints: Pain point quotes sorted by topic and severity (cited from users, without fabricated content).
    • opportunities: A list of opportunity points sorted by signal strength.

Technical Details

The plugin’s core logic is deliberately designed to not depend on the @deepseek-ai/* runtime environment, and is written in pure JavaScript. This means the synthesis engine can be independently tested and maintained outside the DSH framework.

  • Version Compatibility: The plugin has locked the @deepseek-ai/dsh version >=0.1.0--rc.7.
  • Testing Coverage: Tests cover scenarios such as sentence chunking, topic detection, pain point identification, persona inference, scoring and sorting, count truncation, and empty input.

Use Cases

This plugin is suitable for the following roles in specific scenarios:

  • Product Managers / Requirement Analysts: After completing multiple user interviews, quickly synthesize “what users are really complaining about.”
  • UX Researchers / Designers: Extract high-frequency pain points from usability testing notes to support design decisions.
  • Customer Support / Customer Success: Organize user complaints and feedback to identify product improvement priorities.

Conclusion

dsh-plugin-user-research does not aim to replace human judgment; instead, it provides a standardized framework that transforms “user quotes” into “product decisions.” If you are using DSH to build agents and need to process large volumes of user research data, this plugin can serve as a powerful addition to your workflow.