The Agent Skills Directory

[SAFE]: The skill provides standard developer guidance and boilerplate code for the Gemma ecosystem. All instructions align with expected development workflows for AI agents.
[EXTERNAL_DOWNLOADS]: The skill references several official and well-known libraries, including transformers, gradio, google-cloud-aiplatform, and mlx-lm. These are legitimate dependencies for the stated purpose. References to model repositories (e.g., under the google/ organization on Hugging Face) and official documentation domains (e.g., gemma.ai) are consistent with official vendor communications.
[DATA_EXPOSURE]: The Vertex AI deployment guide appropriately recommends using environment variables (GOOGLE_CLOUD_PROJECT, GOOGLE_CLOUD_LOCATION) for configuration rather than hardcoding credentials. This follows security best practices for cloud resource management.
[INDIRECT_PROMPT_INJECTION]: The provided Python and JavaScript templates for chat interfaces ingest user input as part of their core functionality. The code uses standard library patterns for message handling and does not expose high-privilege execution surfaces or unsafe interpolation points.

gemma-dev