Toutiao Publish

自动发布内容到今日头条（微头条/文章）。触发词：发头条、发布头条、微头条、今日头条、发文章、写头条。支持 AI 推荐图片插入正文、免费正版图片库封面、完整文章自动化发布。

This skill automates publishing by running JavaScript inside your logged-in Toutiao browser session and by invoking a local 'browser' CLI and optional local HTTP server. Before installing: (1) verify you trust the author and read the included scripts (publish-toutiao.sh, test-publish.sh) line-by-line; (2) confirm you have the required tooling (a CDP-capable Chrome/Chromium and the 'browser' CLI) — the registry metadata does not list these dependencies; (3) run tests in a separate, disposable browser profile or isolated environment to avoid publishing accidental content or exposing sensitive cookies; (4) don't give this skill access to sensitive local files or shared browser profiles; and (5) if you cannot audit the code yourself, treat it as risky because evaluate() calls can be modified to exfiltrate data visible to the page. If those concerns are acceptable and you audit the scripts, the skill's actions are consistent with its stated purpose.

Type: OpenClaw Skill Name: toutiao-publish Version: 6.1.0 The skill is classified as suspicious due to the broad `fileRead: true` permission declared in `SKILL.md`, which allows reading any local file without explicit justification in the provided scripts for the current version. While `fileWrite`, `network`, and `shell` permissions are restricted to specific paths/domains/commands (`/tmp/openclaw/uploads/`, `mp.toutiao.com + localhost:8000`, `python3 http.server` respectively), the unrestricted `fileRead` capability, combined with the presence of `shell` and `localhost` network access, introduces a potential risk for unauthorized data access, even without clear evidence of intentional malicious exploitation in the current code. The `publish-toutiao.sh` script also directly injects user-provided HTML content into the browser via `innerHTML`, which could lead to XSS on the target platform if not properly sanitized by Toutiao, though this is a vulnerability of the target platform rather than malicious intent by the skill itself.