The Agent Skills Directory

[SAFE]: The skill is authored by GoogleCloudPlatform and uses official Google Cloud SDKs and CLI tools for model tuning and deployment. All external resources and API calls are directed towards established Google Cloud services.
[COMMAND_EXECUTION]: The skill instructs the agent to run gcloud CLI commands for project configuration, authentication, and API enablement (e.g., aiplatform.googleapis.com, storage.googleapis.com). These are standard administrative operations within the Google Cloud ecosystem.
[EXTERNAL_DOWNLOADS]: Verified Python dependencies, including google-cloud-aiplatform, numpy, pandas, and datasets, are installed from standard package registries.
[PROMPT_INJECTION]: The skill includes a surface for indirect prompt injection as it processes user-provided datasets (CSV, JSON, Parquet) for model tuning. This is a primary function of the skill and does not impact the agent's immediate security. Ingestion points: scripts/prepare_dataset.py (line 103) loads external data files. Boundary markers: No explicit delimiters or instructions to ignore embedded prompts are implemented in the formatting scripts. Capability inventory: scripts/tune_model.py submits tuning jobs and scripts/deploy_model.py deploys models to endpoints. Sanitization: scripts/prepare_dataset.py performs basic structural validation and checks for empty or invalid values.

vertex-tuning