The Agent Skills Directory

[DATA_EXFILTRATION]: The scripts/generate_image.py script transmits prompts and configuration data to external third-party services (api.atlascloud.ai and api.nanobananaapi.ai) to generate images. It also attempts to read sensitive API keys from environment variables and a hardcoded local file path (/Users/lifcc/Desktop/code/work/mutil-om/om-generator/.env), which exposes the author's local directory structure.
[COMMAND_EXECUTION]: The skill frequently uses the Bash tool to execute local Python scripts (feed_database.py, generate_image.py) and complex command-line operations. This includes running Google Chrome in headless mode for screenshots and executing inline Python snippets that use PIL and numpy for image processing.
[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves fetching competitor post details (titles, content, analysis) directly from the Xiaohongshu platform to guide its own generation process. This creates a surface for indirect prompt injection where malicious instructions embedded in web content could influence the agent's output or behavior.
[EXTERNAL_DOWNLOADS]: The skill downloads images from URLs provided by the image generation APIs and saves them to the local filesystem using the requests library.

xiaohongshu