The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides technical patterns for using the official google-genai Python library to interact with Vertex AI. All code examples use standard SDK methods without invoking arbitrary shell commands or exposing command injection surfaces.\n- [EXTERNAL_DOWNLOADS]: The skill references standard, well-known software packages including google-genai and pydantic. All external documentation links point to official domains for Google Cloud, Claude, and Vercel.\n- [DATA_EXFILTRATION]: No exfiltration patterns were detected. The skill demonstrates the correct use of environment variables for configuring project IDs and geographic locations, and it does not access sensitive local files or unauthorized network endpoints.\n- [PROMPT_INJECTION]: The instructions do not contain attempts to override agent constraints or bypass safety guidelines. On the contrary, the skill includes a dedicated 'Safety Settings' section as a required pattern, ensuring that harm categories are explicitly blocked in production environments.

vertex-ai-sdk