Introduction¶

Updating a resume on BOSS Zhipin usually involves manual labor: logging into the browser, opening the online resume, and filling out forms section by section. For an agent, there are two main challenges: document parsing (PDF, DOCX, HTML, and Markdown each require different extraction methods) and browser operations (requiring a browser context with a logged-in user). Existing general-purpose browser control plugins (Playwright MCP, Chrome DevTools MCP, browser-use) solve general control, not the narrow path of “synchronizing resumes.”

The 520mxj/boss-zhipin-resume-sync introduced below does just one thing: parse a local resume document, generate structured JSON, and then sync the changes to the user’s logged-in BOSS Zhipin online resume via Microsoft Edge. The plugin is maintained by 520mxj, categorized as a network tool, with 1 star on GitHub. The repository is a standard npm/Cordis package, and the same core skill provides adapter lists for Codex, Claude Code, and DSH simultaneously.

Positioning & Problems Solved¶

The plugin does not hardcode fields for a specific resume. Instead, it extracts the document into a reviewable structured data set before writing it to the online resume. It has several clear boundaries:

1. By default, existing project experience is preserved unless the user explicitly requests modification or deletion.
2. It does not take over the login or bypass CAPTCHAs; when encountering security verification, it waits for the user to complete it manually.
3. It does not save account passwords and deletes the temporary Edge user directory after synchronization ends.
4. General browser control capabilities are not reimplemented; the repository retains only the narrow-range, domain-restricted CDP bridge required for BOSS resume synchronization. See docs/duplicate-check.md for the deduplication basis.

Supported Inputs¶

Fields that cannot be recognized are not hallucinated; missing information is exposed to the user for confirmation as-is.

Workflow¶

The workflow consists of six steps. The first three are completed locally for parsing and validation, while the latter three handle the browser:

1. extract reads the uploaded document.
2. The Agent generates resume.json based on the Resume schema (skills/boss-zhipin-resume-sync/references/resume.schema.json).
3. validate validates the structured resume.
4. start copies necessary Edge login data to a temporary directory and opens the BOSS online resume using CDP listening only on 127.0.0.1.
5. The Agent uses read, fill, click, and eval to save sections block by block and review items one by one.
6. stop closes the debug browser and deletes the temporary login copy.

You need to close all ordinary Edge windows before executing start. The script will not automatically close other browser windows to avoid losing user data.

Installation & Enablement¶

First, install dependencies:

npm install

The DSH-side installation command comes from the original README and does not require git clone:

dsh plugin add github:520mxj/boss-zhipin-resume-sync

The repository contains cordis.patch.yml and lib/index.js, which is a standard npm/Cordis package. DSH registers the same Skill via ctx.skills.registerProvider(). The author declares peerDependencies in package.json: @deepseek-ai/cordis ~4.0.4 and @deepseek-ai/dsh-skill ^0.2.0-rc.2, with engines requiring Node >=20.16. There is also bin: boss-resume pointing to skills/boss-zhipin-resume-sync/scripts/boss-resume.mjs.

Installation for Codex and Claude Code is different: Codex looks at the root directory .codex-plugin/plugin.json and the standard skills/ folder, while Claude Code looks at .claude-plugin/plugin.json; both can also simply place the skills/boss-zhipin-resume-sync directory into a project or personal skills directory.

Typical Usage¶

CLI quick verification goes through this chain; the path needs to match the plugin installation location:

node skills/boss-zhipin-resume-sync/scripts/boss-resume.mjs extract ./resume.pdf --out ./.boss-resume/source.json
node skills/boss-zhipin-resume-sync/scripts/boss-resume.mjs validate ./.boss-resume/resume.json
node skills/boss-zhipin-resume-sync/scripts/boss-resume.mjs start
node skills/boss-zhipin-resume-sync/scripts/boss-resume.mjs read --max-chars 20000
node skills/boss-zhipin-resume-sync/scripts/boss-resume.mjs stop

The repository comes with a development verification script that can be used to check if the installation is complete:

npm test
node skills/boss-zhipin-resume-sync/scripts/boss-resume.mjs --help

Applicable Scenarios & Notes¶

Suitable for developers who need to sync their local resume documents to a BOSS Zhipin online resume while wanting a verifiable process. After fields are generated, they are first persisted as resume.json and then written into the browser, facilitating item-by-item verification.

Several notes:

1. The plugin runs with the current DSH process permissions. Before installing, you should read the source code and license (the repository states MIT) to confirm that peerDependencies match the local versions before enabling.
2. Security boundaries are limited to www.zhipin.com: the CLI’s eval, fill, and click are only allowed to execute against this domain.
3. The plugin does not handle passwords, bank cards, CAPTCHAs, or authentication popups; explicit authorization must be obtained from the user before deleting resume blocks or contact information.
4. The temporary browser profile directory is created, used, and cleaned up by the script; do not commit it to Git.
5. This tool serves the user’s own job application resume and does not include automatic application submission, batch scraping, recruiter actions, or risk control evasion features.

Ending¶

The value of this plugin lies in splitting “document parsing + browser resume writing” into a narrow path while simultaneously providing adapter lists for three sets of Agents. The directory page is at https://www.skillhub.cn/plugins/520mxj/boss-zhipin-resume-sync , and the source code is at https://github.com/520mxj/boss-zhipin-resume-sync . Please verify the source code, license, and dependency versions yourself before using.