The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/publish.py executes system browser binaries (Chrome or Edge) using subprocess.Popen with the --remote-debugging-port flag to enable automated interactions.
[CREDENTIALS_UNSAFE]: Authentication cookies are captured during the login process and stored in a local .zhihu_cookies.json file. A persistent browser profile is also created in the .zhihu_browser_profile directory.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of Playwright and its associated browser binaries from external repositories to function.
[DATA_EXFILTRATION]: Local article content and authentication metadata are transmitted to Zhihu's official API endpoints and web services during the publishing workflow.
[PROMPT_INJECTION]: The skill processes user-supplied Markdown files without sanitization, creating an indirect prompt injection surface. \n
Ingestion points: Markdown files read by convert.py and publish.py. \n
Boundary markers: None identified; content is processed without delimiters. \n
Capability inventory: Local command execution, file system access, and network communication with Zhihu APIs. \n
Sanitization: No validation or cleaning is performed on input Markdown before it is converted to HTML and submitted to the target platform.

md-to-zhihu