The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection because it ingests untrusted data from arbitrary URLs using the baoyu-url-to-markdown module. This content is subsequently used as context for formatting, summarization, and image generation prompts, which could allow maliciously crafted web pages to influence the agent's behavior.
[EXTERNAL_DOWNLOADS]: The skill depends on external system software and libraries, specifically the Chrome browser, Node.js, and the Python packages requests and Pillow. These are standard tools for the described tasks of web scraping and image processing.
[CREDENTIALS_UNSAFE]: The skill handles sensitive account information for automated publishing on WeChat and X. It recommends storing API keys in a local .env file and utilizes browser session data for automation. While these are necessary for the skill's purpose, users should be aware that any agent capability with network access can technically access these local configuration files if not properly restricted by the platform.
[COMMAND_EXECUTION]: The skill utilizes automated browser control (Chrome CDP) to navigate login pages and publishing interfaces, explicitly mentioning the capability to bypass anti-automation detection mechanisms on platforms like X/Twitter.

content-creation-publisher