Introduction

The core philosophy of DeepSeek Harness (DSH) is “everything is a plugin.” For developers preparing for technical interviews, having an interviewer that can simulate real pressure and follow up layer by layer from your resume is essential. Most existing plugins focus on resume generation or hiring assistance, lacking deep simulation of the “interview process” itself.

The dsh-interviewer plugin is born from this need. It uses the model already configured in DSH to act as a technical interviewer. It does not explain, correct, or encourage; it only tests and measures the candidate’s true capability boundary within a limited time.

Plugin Overview

Plugin Name: dsh-interviewer
Maintainer: ufire
Category: Model Inference
License: MIT
GitHub: ufiredong/dsh-interviewer
Community Directory: skillhub.cn/plugins/ufiredong/dsh-interviewer

This is a tool that integrates a “mock interview skill” with a “draggable panel UI.” It is resume-driven, asks progressive follow-up questions (L0-L5), scores evidence-based performance across five dimensions and derives a salary range. It is registered as a skill and also provides a panel plugin.

Installation and Activation

Run the following command in the terminal to install:

dsh plugin --profile web add github:ufiredong/dsh-interviewer

After installation, you must restart dsh web for the plugin to take effect. After restarting, a floating “Mock Interview” button will appear in the bottom-right corner.

Core Features

The plugin provides the following capabilities:

  • Real model as interviewer: Uses the model already configured in DSH and does not require an additional API key.
  • Resume-driven progressive follow-up: Starts from resume facts and uses L0-L5 staged questioning to cover fact verification, technology selection, implementation details, counterfactual scenarios, and transferability.
  • Multiple styles × multiple rounds: Supports four interviewer styles (gentle guidance, pragmatic engineering, cold stress-test, and academic rigor) and four interview rounds (first technical, second technical, hiring manager, and HR).
  • Built-in state machine: Prevents getting stuck. Each topic can be followed up on at most twice to ensure assessment breadth.
  • Evidence-based scoring: Five weighted dimensions, where every point scored must be traceable to the candidate’s original statement. The salary range is calculated by a deterministic script, not estimated by the model.
  • Visible process: The status bar at the top of the panel shows elapsed time, interview plan progress, and the number of follow-up questions for the current topic.
  • Voice responses: Supports the built-in speech recognition in browsers (Chrome/Edge).
  • Strict mode: Does not explain, correct, or encourage; only tests.

Usage

  1. Open settings: Click the floating button in the bottom-right corner to open the settings dialog.
  2. Configure parameters: Select the target area (such as LLM/Agent), job level, interview round, and interviewer style.
  3. Start the interview: After confirmation, enter the interview panel. Paste your resume into the chat box (PDF reading is not supported), or simply say “Start the interview.”
  4. Invoke from the main session: After installation, the plugin automatically registers a skill named interviewer, so you can directly use the skill interviewer command in the main session.
  5. Scoring report: After the interview ends, the scoring report must be generated by running a script in the main session; it is not displayed inside the panel.

Scoring and Salary

The plugin scores candidates using a five-dimension weighted model. The salary range is calculated by the skills/interviewer/scripts/score.mjs script.

Scoring Dimensions:
* Technical Depth (30%)
* System Design (25%)
* Engineering Rigor (20%)
* Expression and Structure (15%)
* Boundary Awareness (10%)

Salary Range Examples (total package in first-tier cities):
* ≥ 4.5: Expert (P8 / 3-1 / T7), annual salary 1.2 – 1.8 million
* 3.8 – 4.5: Senior (P7 / 2-2 / T6), annual salary 700,000 – 1.2 million
* 3.0 – 3.8: Mid-level (P6 / 2-1 / T5), annual salary 450,000 – 700,000
* 2.2 – 3.0: Junior (P5 / 1-2 / T4), annual salary 300,000 – 450,000
* < 2.2: Not recommended to proceed

Notes

  1. Privacy and Security: Your resume and all conversation content are sent to the model service provider configured in DSH. Do not paste sensitive information.
  2. History: Interview conversations do not enter the DSH conversation history. State is stored in browser memory and will be lost after restarting the session or browser.
  3. PDF Support: Only Markdown or plain text pasting is supported; PDF files cannot be read directly.
  4. System Requirements: Node.js >= 20 is required.

Conclusion

dsh-interviewer addresses two core problems in mock interviews: “how to follow up” and “how to evaluate quantitatively.” Through an explicit state machine design, it avoids the model getting infinitely stuck on a single topic and ensures the breadth and efficiency of the interview. If you are looking for a practice method that can genuinely reflect your technical boundary capabilities, this plugin is a reliable tool.