The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/gzh_archive.py invokes shell commands via subprocess.check_call to automate the installation of Python dependencies and uses subprocess.run to call ripgrep (rg) for performing full-text searches on the archived Markdown files.
[DYNAMIC_EXECUTION]: The skill uses __import__ to dynamically check for the presence of required Python modules at runtime. It also utilizes Selenium's execute_script and execute_cdp_cmd methods to inject custom JavaScript into the browser environment to facilitate navigation, bypass bot detection, and extract data.
[EXTERNAL_DOWNLOADS]: The scraper fetches article content and images from external domains associated with WeChat and Sogou. It also initiates the download of standard Python libraries from the public package registry to ensure environment readiness.
[INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted content from web pages and converts it to Markdown for use within the agent's context. While the skill applies sanitization—removing scripts, styles, and iframes—the ingestion of third-party data represents an indirect injection surface.
[CREDENTIALS_UNSAFE]: To maintain user login states, the skill persists browser profile data, including session cookies, in the local directory ~/.opencode-gzh-profile. While intended for user convenience, access to this directory could expose active session credentials.

gong-zhong-hao