Skip to content

Guide · Agent skill

How to use the Zhihu Article & Video Publisher skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The Zhihu Article & Video Publisher skill (zhihu-upload) lets an AI agent publish articles and single videos to Zhihu, a Chinese Q&A and publishing platform. It works by calling the sau command-line tool from the social-auto-upload project rather than writing its own upload code. This guide explains what the skill does, how to install it, how to ask for it in plain language, and what to check before anything goes live.

Zhihu Article & Video Publisherdreammis/social-auto-upload · MIT

Quick start

  1. 1Get the skill files. Download the skill ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder. You can also upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
  2. 2Install and prepare sau. Make sure the social-auto-upload project and its sau command-line tool are installed, following that project's documentation.
  3. 3Log in to Zhihu. Run sau zhihu login --account <name> and enter your password and any verification code in the login window. Then run sau zhihu check to confirm the session matches the account.
  4. 4Ask the agent for a preview. Request the publish with --prepare-only so the agent fills the form and checks it without going live.
  5. 5Review the preview. Check the text, image order, links, category and declaration yourself before approving a real publish.
  6. 6Publish and record the result. Remove --prepare-only to publish. Read the status in the result file: published means confirmed, and a url is then available.
  7. 7Verify the work. Run sau zhihu verify with the result file to open the work link and confirm the title matches.

What the skill lets the agent do

Once installed, the agent can run five Zhihu actions through the sau zhihu commands: log in, check a saved session, upload an article, upload a video, and verify a published work by opening its link.

Articles can be HTML, Markdown or TXT. Images stay in the position they hold in your source file. Videos are single files, with a title, description, originality flag, category and content declaration.

  • Login: sau zhihu login --account <name> saves a local session for an account alias.
  • Account check: sau zhihu check --account <name> confirms the session matches the current account.
  • Article upload: sau zhihu upload-article for HTML, Markdown or TXT sources.
  • Video upload: sau zhihu upload-video for one video file.
  • Verify: sau zhihu verify --account <name> --result <json> opens the work link in a result file and checks the title.

Install the skill by ZIP or Claude app upload

Download the skill ZIP from this page and unzip it. For a personal install, place the skill folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.

You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.

  • Personal install: unzip into ~/.claude/skills/
  • Project install: unzip into the project's .claude/skills/ folder
  • Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Prepare the sau tool and a Zhihu login

The skill cannot publish on its own. It needs the social-auto-upload project with the sau command-line tool installed, and a Zhihu account with a locally saved login session. The source does not give installation steps for sau, so follow the project's own documentation for that part.

Log in with the account alias you want to use. Enter your password and any verification code in the login window yourself. The session file is saved on your machine and is not committed to the repository. The --account value is a local alias and does not need to match your Zhihu display name.

  • Requires: sau installed and on your path
  • Requires: a Zhihu account with a saved local session
  • Example login: sau zhihu login --account creator
  • Example check: sau zhihu check --account creator
sau zhihu login --account creator sau zhihu check --account creator

Publish an article

Article titles can be up to 100 characters. Put images in the body with relative paths. Relative paths resolve from the folder containing the source file. For Markdown, run sau article setup first, the first time you use it.

Before submitting, check the text, image order, basic formatting and links. If the editor restores a non-empty draft, the upload stops so existing content is not overwritten. Custom theme CSS, manual covers, column inclusion, Zhihu's cheese tasks and document import are not supported.

  • Use UTF-8 HTML, Markdown or TXT
  • Image paths are relative to the source file's folder
  • Uploads stop if the editor has a non-empty restored draft
sau zhihu upload-article --account creator --title "文章标题" --content-file "文章.md"

Publish a video with the required options

Video files must be larger than 100 KB (102,400 bytes), no more than 20 GB, and no longer than 4 hours. The title must be 5 to 50 characters and the description can be up to 300 characters. The platform's generated cover is used by default.

Pass --originality as original or reprint. Pass --category for the category, or see sau zhihu upload-video --help for the options. If you omit --category, the command shows a live menu for you to choose from by number. For non-interactive runs, specify the category and declaration. The program still checks the live menu before submitting.

Use --video-mark with one of the listed declarations: 内容无需标注, 含 AI 生成内容, 含虚构演绎内容, 内容含营销信息, 内容为转载, or 个人观点仅供参考. The declaration must match the actual content.

  • --originality: original or reprint
  • --category: pick from the help output or the live menu
  • --video-mark: must accurately describe the content
  • Use --show-options to fetch current category and declaration choices without publishing
sau zhihu upload-video --account creator --file "视频.mp4" --title "视频标题示例" --desc "视频简介" --originality original --category "教育" --video-mark "含 AI 生成内容"

Preview first, then check the result

Both articles and videos accept --prepare-only, which fills in the form and checks it, then stops without publishing. Be aware that --prepare-only and --show-options can both leave a draft on Zhihu's side.

A published status means the platform confirmed publication, and the detail link is written to the url field of the result file. If detail loading fails, the result keeps published and error. Check the Creator Center for the original work and do not resubmit. A prepared status does not mean the work was published. If the status is submission_unknown, check the original work by title and submission time before trying again. Once a URL exists, run verify.

  • Use --prepare-only to review before going live
  • published: confirmed; read url from the result file
  • prepared: not published
  • submission_unknown: check the Creator Center by title and time first
  • Never retry blindly, since this can create duplicate posts
sau zhihu verify --account creator --result "本次结果.json"

Scheduling and other limits

The --schedule option takes a Beijing-time value in the form YYYY-MM-DD HH:MM. It must be between 30 minutes and 48 hours from now. It is marked experimental, and the full scheduling flow is not guaranteed to work, so do not rely on it for unattended publishing.

The source also lists these as unsupported for the skill: automatic topic selection, custom covers, columns and content source selection.

  • --schedule is experimental and not guaranteed
  • Scheduling window: 30 minutes to 48 hours ahead, Beijing time
  • Unsupported: custom covers, columns, topic selection, content source

What people use it for

  • Re-posting a Markdown article with local images to Zhihu while keeping the image order intact.
  • Uploading a single finished video with a title, description, originality flag and content declaration.
  • Previewing a batch of article submissions with --prepare-only before publishing any of them.
  • Checking that a saved Zhihu login still belongs to the expected account before a publishing run.
  • Confirming a published work's title by opening its link from the result file.

Frequently asked questions

How do I install the zhihu-upload agent skill?

Download the skill ZIP from this page and unzip it into ~/.claude/skills/ for a personal install, or into a project's .claude/skills/ folder for one project. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.

What do I need before the skill can publish to Zhihu?

You need the social-auto-upload project with the sau command-line tool installed, and a Zhihu account with a locally saved login session. The skill cannot publish without both.

Is the Zhihu Article & Video Publisher skill free?

The listing states the licence as MIT, while the SKILL.md header names MIT-0. Check the licence text in the downloaded files for the exact terms before reusing it.

Does the skill enter my Zhihu password for me?

No. Your password and verification codes are entered by you in the login window. The session file is stored locally and is not committed to the repository.

What should I do if an upload status is submission_unknown?

Do not resubmit. Check the Creator Center for the original work by its title and submission time first, because a retry could create a duplicate post.

Can the skill schedule Zhihu posts to publish later?

The --schedule option exists but is marked experimental, and the full scheduling flow is not guaranteed to work. Avoid relying on it for unattended publishing.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits Chinese-language creators and teams who already use social-auto-upload and want an agent to post articles and single videos to Zhihu, ideally with a --prepare-only review step. It is a poor fit if you need custom covers, columns, topic selection, document import or reliable scheduled publishing, since the source says these are unsupported or experimental.

Download the skill