The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes curl commands and local Python scripts to interact with search APIs and process results. It mitigates injection risks by using Python's urllib.parse.quote and json.dumps to sanitize user input before it is passed to the shell.
[EXTERNAL_DOWNLOADS]: The skill interacts with established search providers (Tavily, Serper, Brave, etc.) and uses the Google Gemini API for image generation. These are well-known or trusted services. The skill also downloads images, which are then validated and filtered for quality using the Pillow library.
[CREDENTIALS_UNSAFE]: Secret management is handled through environment variables or .env files, which is the recommended approach for local tools. The skill instructions explicitly warn against printing raw API keys and provide masking examples.
[DATA_EXFILTRATION]: No unauthorized data transmission was found. Network activity is limited to the search and generation APIs necessary for the skill's documented functionality.
[SAFE]: The skill's behavior is entirely consistent with its description. It implements comprehensive query optimization, result cleaning, and error handling, demonstrating a high level of engineering quality and security awareness.

web-search