Introduction

In the plugin scenario of DeepSeek Harness (DSH), the job hunting process is not merely about generating a response: local job files need to be organized, resumes need to be parsed, jobs seen in the browser need to be marked, and finally, a searchable list of interested jobs and reports need to be formed.

The dsh-job-hunting introduced below is a local-first DSH plugin and Runtime Skill used to organize job postings, confirm job profiles, and maintain interest lists.

User resumes, confirmed profiles, interest data, reports, and generated websites are only written to the active Workspace selected by the host; the plugin does not use a fixed personal data directory.

What is this

dsh-job-hunting is a DeepSeek Harness plugin and Runtime Skill, with the repository address at allentnetus/dsh-job-hunting.

Basic information is as follows:

  • Plugin Name: dsh-job-hunting
  • License: v0.1 own code and templates use the MIT license; third-party dependencies retain their respective licenses
  • Verification Scope: The following examples target the DeepSeek Harness 0.1.0-rc.6 web profile
  • Repository Address: https://github.com/allentnetus/dsh-job-hunting

Core Capabilities

The main capabilities of v0.1 are introduced below.

  1. Import local JSON or local Markdown job files, and complete standardization, deduplication, matching, and report generation.

  2. Parse DOCX, text-based PDF, TXT, and Markdown resumes; scanned or encrypted PDFs will be explicitly rejected, v0.1 does not use OCR.

  3. Build a JH / Job Hunting Intelligence Station / Job Hunting static website. Data is embedded during the build phase, and the generated index.html can be opened via file://.

  4. Job marking in the browser is only saved in the current browser’s temporary state; users can export it and then sync it to the formal interest ledger in the active Workspace.

  5. Register job-hunting Runtime Skill and job_hunting_ tools.

  6. Directly register and enable the job_hunting_collect_browser_jobs tool by default, reading structured jobs visible in whitelist sites via Tencent/BrowserSkill.

  7. Provide module-level Windows shortcut services. It does not automatically create shortcuts; the shortcut user entry is not provided until the host is integrated.

Prerequisites

First, ensure that Node.js, pnpm, and the DSH command are already installed on the local machine. You can check them using the following commands:

node --version
pnpm.cmd --version
dsh.cmd --help

Prerequisites include:

  • DeepSeek Harness 0.1.0-rc.6
  • Node.js >=24.15.0
  • pnpm >=11.19.0
  • A target profile that is initialized or can be initialized; the example uses web, other profiles need to replace web in the command.

If pnpm is not installed yet, you can install a specific version globally:

npm.cmd install --global pnpm@11.19.0

If DSH CLI is not installed yet, you can install a verified version globally:

npm.cmd install --global @deepseek-ai/dsh@0.1.0-rc.6
dsh.cmd --help

If you only want to run it temporarily, you can also use npx:

npx.cmd @deepseek-ai/dsh@0.1.0-rc.6 --help

Installing the Plugin

The following installs v0.1.3 using the GitHub tag tarball:

dsh.cmd plugin --profile web add "https://codeload.github.com/allentnetus/dsh-job-hunting/tar.gz/refs/tags/v0.1.3"

After installation, check the plugin list and confirm that relevant entries appear in the configuration:

dsh.cmd plugin --profile web list
dsh.cmd --profile web --dump-config | Select-String 'dsh-job-hunting|job-hunting'

Updates and Rollbacks

When updating, check first, then explicitly execute the update:

dsh.cmd plugin --profile web outdated
dsh.cmd plugin --profile web update dsh-job-hunting

After updating, you can check the configuration again:

dsh.cmd --profile web --dump-config | Select-String 'dsh-job-hunting|job-hunting'

Silent automatic upgrades are not performed by default. If an update requires a rollback, restore the profile’s package.json and pnpm-lock.yaml first, then execute:

dsh.cmd plugin --profile web install --frozen-lockfile

If pnpm prompts that build scripts for Git dependencies need approval, you can allow this package in your own DSH profile’s pnpm-workspace.yaml, then reinstall:

allowBuilds:
  dsh-job-hunting: true

Browser Job Scraping

The job_hunting_collect_browser_jobs tool depends on Tencent/BrowserSkill. The bsk CLI and browser extension here are external prerequisites that are not installed or published with this plugin package.

Before using browser scraping, first check if BrowserSkill is available:

bsk --help
bsk status

This tool only reads structured jobs visible in whitelist sites. When calling, confirmed: true must be passed, and it is limited by maxItemsPerRun and minIntervalMs.

The scraping process does not extract credentials, submit applications, send recruitment messages, or bypass authentication or CAPTCHA controls.

Scheduled Reminders

schedule.enabled in the plugin configuration is false by default. The plugin does not create an independent background scheduler.

When using the native DSH Schedule, apply the host-level overlay explicitly:

dsh.cmd web --patch $overlayPath

Native DSH Schedule is only valid for the current session. New sessions need to use --patch again and recreate the reminders.

After a reminder expires, the reminder itself does not constitute scraping authorization. The user must confirm the specific whitelist URLs and read-only scope for this round before calling job_hunting_collect_browser_jobs.

Local Check Commands

If you need to check the plugin code locally, you can use the following commands:

pnpm.cmd install
pnpm.cmd test
pnpm.cmd typecheck
pnpm.cmd build
pnpm.cmd lint
pnpm.cmd dsh:smoke
pnpm.cmd release:check

Use Cases and Notes

The points suitable for use and points that need special attention are listed below.

Suitable users:

  • Need to organize local JSON or Markdown job files;
  • Need to parse DOCX, text-based PDF, TXT, or Markdown resumes;
  • Need to build a static job site that can be opened via file://;
  • Need to maintain interest lists and ledgers in DSH conversations;
  • Need to read structured jobs visible in the browser on whitelist sites.

Notes:

  • The plugin will register Runtime Skills and tools, and tools can be called within DSH sessions; source code, configuration items, licenses, and external dependencies should be checked before enabling.
  • The plugin runs with the permissions of the current DSH process; before enabling, confirm the local environment, external CLI, and browser session permission scopes.
  • Data is written to the active Workspace selected by the host, not a fixed personal data directory.
  • Job marking in the browser is a temporary state; the formal ledger needs to be synced after export.
  • v0.1 does not use OCR; scanned or encrypted PDFs will be rejected.
  • The plugin will not automatically create Windows shortcuts; the shortcut user entry is not provided until the host is integrated.
  • Scheduled reminders are not a background scheduler, nor do they grant unmanned scraping authorization.

Conclusion

dsh-job-hunting provides a local-first job hunting process: from job file import, resume parsing, interest list maintenance, to static website generation and whitelist browser job scraping.

GitHub Repository Address: https://github.com/allentnetus/dsh-job-hunting