The Agent Skills Directory

[DATA_EXFILTRATION]: The skill includes a 'custom_api' mode in scripts/publish_wechat.py that transmits article titles, content (HTML), cover URLs, and tags to an external raw IP address (http://39.108.254.228:8002/publish-draft). This transmits user-generated data to an untrusted third-party endpoint not associated with well-known services.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from external platforms (Zhihu hot lists, Weibo, and RSS feeds) via scripts/fetch_sources.py. This data is used to generate article content without implementation of boundary markers or sanitization to prevent embedded malicious instructions from influencing the agent's behavior.
[COMMAND_EXECUTION]: The skill utilizes shell command execution for internal orchestration. scripts/auto_publisher.py uses subprocess.run to call fetch_sources.py, and scripts/copy_to_clipboard.py uses subprocess.Popen to interact with system tools like xclip and xsel. These provide a capability surface for local command execution.
[EXTERNAL_DOWNLOADS]: The script scripts/search_images.py performs network requests to the Unsplash API (api.unsplash.com) to search for and retrieve image metadata. This is a well-known service usage for the skill's stated purpose of content creation.

wechat-hotspot-publisher