How do I install the Social Capture skill?

Run npx skillmds@latest add yang0/social-capture in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Social Capture safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Social Capture?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Social Capture free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Social Capture?

yang0 (@yang0) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Social Capture by yang0 — Agent Skill

Name: Social Capture
Author: yang0

File contents

social-capture skill

Use this skill when an AI needs an evidence image of a specific social post, answer, note, tweet, or Douyin profile. The project is screenshot-only: it does not search, publish, reply, download original media, or bypass CAPTCHAs.

Workflow

Ask for or identify one or more canonical content URLs. Do not invent a URL.
Ask the user for an explicit output directory. Never use the repository or a default downloads folder.
Check capabilities with social-capture providers and environment health with social-capture doctor --json.
If search is explicitly requested by the user, run the read-only social-capture search-backends --json command, check whether the hinted backend already exists, and install only a missing backend using its upstream instructions. The AI may perform that explicitly requested installation and then run the backend; social-capture itself never silently installs, clones, or runs third-party projects. Read and follow each backend's license, authentication, and platform terms first.
Capture with social-capture capture URL --output-dir DIR. Use --platform when auto-detection is ambiguous.
Read manifest.json, verify status, dimensions, and SHA-256 before using an image in content. Report blocked, deleted, or partial items honestly.

Examples

social-capture capture "https://x.com/user/status/123" --output-dir "G:\captures\x"
social-capture capture "https://www.zhihu.com/question/1/answer/2" --output-dir "G:\captures\zhihu"
social-capture capture "https://www.xiaohongshu.com/explore/ID?xsec_token=..." --output-dir "G:\captures\xhs" --xhs-all-images
social-capture capture "https://www.douyin.com/user/SEC_UID" --output-dir "G:\captures\douyin" --douyin-rows 2

For Xiaohongshu, signed xiaohongshu.com URLs must include xsec_token; short xhslink.com URLs are accepted and resolved by Chrome. Use --xhs-comments 0 to hide comments; the default is eight comments.

Search routing hints

Search is an optional external action and must be separately authorized by the user. The current read-only hints are:

Weibo: https://github.com/dataabc/weibo-search
X: https://github.com/public-clis/twitter-cli (Apache-2.0)
Xiaohongshu: https://github.com/xpzouying/xiaohongshu-mcp (Apache-2.0)
Douyin/Zhihu: https://github.com/NanmiCoder/MediaCrawler (Other; read its license first)

Agent-Reach can be an optional routing layer for X/Xiaohongshu in workflows that already have it installed. None of these projects is a social-capture dependency.

Safety

Do not expose Cookie headers, xsec_token, signed URLs, or temporary profile paths in generated prose. A screenshot must be a real content-card crop; if a Provider cannot uniquely locate the target, preserve the error instead of capturing a feed or full page.

---name: social-capturedescription: Capture a focused screenshot card from Weibo, Zhihu, X, Xiaohongshu or Douyin through a logged-in local Chrome CDP session.---# social-capture skillUse this skill when an AI needs an evidence image of a specific social post,answer, note, tweet, or Douyin profile. The project is screenshot-only: itdoes not search, publish, reply, download original media, or bypass CAPTCHAs.## Workflow1. Ask for or identify one or more canonical content URLs. Do not invent a URL.2. Ask the user for an explicit output directory. Never use the repository or a default `downloads` folder.3. Check capabilities with `social-capture providers` and environment health with `social-capture doctor --json`.4. If search is explicitly requested by the user, run the read-only `social-capture search-backends --json` command, check whether the hinted backend already exists, and install only a missing backend using its upstream instructions. The AI may perform that explicitly requested installation and then run the backend; social-capture itself never silently installs, clones, or runs third-party projects. Read and follow each backend's license, authentication, and platform terms first.5. Capture with `social-capture capture URL --output-dir DIR`. Use `--platform` when auto-detection is ambiguous.6. Read `manifest.json`, verify `status`, dimensions, and SHA-256 before using an image in content. Report blocked, deleted, or partial items honestly.## Examples```powershellsocial-capture capture "https://x.com/user/status/123" --output-dir "G:\captures\x"social-capture capture "https://www.zhihu.com/question/1/answer/2" --output-dir "G:\captures\zhihu"social-capture capture "https://www.xiaohongshu.com/explore/ID?xsec_token=..." --output-dir "G:\captures\xhs" --xhs-all-imagessocial-capture capture "https://www.douyin.com/user/SEC_UID" --output-dir "G:\captures\douyin" --douyin-rows 2```For Xiaohongshu, signed `xiaohongshu.com` URLs must include `xsec_token`;short `xhslink.com` URLs are accepted and resolved by Chrome. Use`--xhs-comments 0` to hide comments; the default is eight comments.## Search routing hintsSearch is an optional external action and must be separately authorized by theuser. The current read-only hints are:- Weibo: `https://github.com/dataabc/weibo-search`- X: `https://github.com/public-clis/twitter-cli` (Apache-2.0)- Xiaohongshu: `https://github.com/xpzouying/xiaohongshu-mcp` (Apache-2.0)- Douyin/Zhihu: `https://github.com/NanmiCoder/MediaCrawler` (`Other`; read its license first)Agent-Reach can be an optional routing layer for X/Xiaohongshu in workflowsthat already have it installed. None of these projects is a social-capturedependency.## SafetyDo not expose Cookie headers, `xsec_token`, signed URLs, or temporary profilepaths in generated prose. A screenshot must be a real content-card crop; if aProvider cannot uniquely locate the target, preserve the error instead ofcapturing a feed or full page.

Frequently asked questions

Social Capture