How to use the xueqiu-upload agent skill for Xueqiu long posts

What is an agent skill, and what does this one do?

An agent skill is a folder of instructions and optional scripts that an AI agent loads when your request matches the skill's description. This one is described in its SKILL.md file and covers Xueqiu long-form posts only.

Once triggered, the agent can log in to a Xueqiu account, check that the login matches the expected user, upload an HTML, Markdown or TXT article, keep in-body images where they appear in your source file, set a separate cover image, and query the published work in the backend to confirm it went through.

• Accepted article formats: HTML, Markdown and TXT.
• Images inside the body stay at their positions in the source file.
• A separate cover image can be set with --cover.
• Not supported: scheduled posting, video, and articles with tables or code blocks.

Install the xueqiu-upload skill

Download the ZIP from the skill page on this site. You can install it in one of two ways.

For Claude Code, unzip the folder into ~/.claude/skills/ for all your projects, or into a project's .claude/skills/ folder for one project only. For the Claude apps, upload the ZIP under Settings → Capabilities → Skills.

The skill calls the sau CLI, so you also need the social-auto-upload repository set up with its sau command installed. The SKILL.md says that on Windows, if sau is not on your PATH, you can run .\.venv\Scripts\sau.exe from the repository root.

• Python and the social-auto-upload repository with the sau CLI installed.
• Google Chrome. Visible mode is the default.
• A Xueqiu creator account that you log in to yourself in the browser.

Log in and check your Xueqiu account

The first time you use an account, the agent runs a login command. A Chrome window opens and you sign in there. The tool saves the session and ties it to your real Xueqiu user ID. Passwords should never be typed into commands or source code.

The login command takes an expected display name, so the tool can confirm it is the right account. If you already have a saved account, the agent runs a check command instead. Check the --help output of the CLI before you run anything, because flags can change.

• First login: sau xueqiu login --account <alias> --expected-name "your nickname"
• Existing account: sau xueqiu check --account <alias>
• Note: a login made in another browser tool does not automatically create the CLI session file.

Prepare your article and choose the options

The agent needs four things: a non-empty single-line title, a UTF-8 content file in HTML, Markdown or TXT, a declaration that matches how the article was made, and a visibility setting. Visibility is either 公开 (public) or 私人 (private), and the default is public.

The declaration must match reality. The skill allows 不添加创作声明 (no creation declaration) or 本文包含AI生成内容 (contains AI-generated content). The agent should ask you rather than guess.

For images, write them in the article where they belong, using HTML img tags or Markdown image syntax. Relative paths are resolved from the article's folder. Images must be local GIF, JPEG or PNG files; remote URLs and data URLs are not accepted. Markdown is preferred. The first time you use Markdown, run sau article setup.

• Title: non-empty, one line.
• Content: HTML, Markdown or TXT, UTF-8.
• Between-paragraph image example: 
• Custom CSS is not guaranteed to be kept. Headings become platform subheadings, and lists become numbered or bulleted paragraphs.

Run a draft first, then publish

The upload command publishes by default. To check the draft and preview without publishing, add --prepare-only. Xueqiu saves a remote draft automatically, so running the same command again can create more than one draft. If the editor already contains content, the tool refuses to overwrite it. Leave out --cover if you do not need a separate cover image.

The example below is taken from the skill. Replace the account alias, title, file names and declaration with your own. Run it only after you have reviewed the options.

• Preview only: add --prepare-only.
• Private post: --visibility 私人.
• Check the CLI help for the current options first: sau xueqiu upload-article --help

Read the result and recover from interruptions

The command saves a result JSON file and screenshots. Keep them. A status of prepared means the draft is ready. A status of published means the platform accepted the post and the tool checked the subject, title, visibility and body. The field management_verified: true means a backend query by the original work ID also passed. A private work shows as visible only to you.

If a result is recorded as submission_unknown, do not re-run the upload or click publish again, since that can create duplicates. Instead, run the verify command with the saved result file. It performs a read-only lookup of the original work ID and compares the backend content and images with your original article. If there is no ID, search 内容管理 (content management) by title, account and time.

• Keep the result JSON path and screenshots.
• Verify: sau xueqiu verify --account <alias> --result "result-file.json"
• Use --debug to save a screenshot when a step fails.
• Visible Chrome is the default. --headless is optional, and the skill notes that headless access was blocked by the platform in testing.
• A network timeout, a 403 error or a subject mismatch should be reported as-is. Do not assume your cookies have expired from that alone.