The Agent Skills Directory

Script Generation and Execution: The skill facilitates prompt management by generating Python scripts that utilize the google-cloud-aiplatform library. These scripts are executed within the agent's environment to perform actions like creating or deleting resources. To mitigate the risk of unintended actions, the skill implements a 'Safety & Confirmation Tier' system that mandates interactive or typed confirmation for any operation that modifies or deletes data.
Indirect Prompt Injection Surface: The skill processes user-defined prompt templates and performs variable interpolation through the assemble_contents method. This represents a potential surface for indirect prompt injection, as malicious instructions could be embedded in the data being processed. However, this functionality is inherent to prompt orchestration, and the skill relies on the underlying SDK's handling of these templates.
Dependency Management: The skill requires the installation of the google-cloud-aiplatform and google-genai Python packages. These are official libraries provided by the service vendor and are used for their intended purpose of interacting with Vertex AI services.
Credential Handling: The skill relies on Google Cloud Application Default Credentials (ADC) configured in the user's environment. It does not store or request hardcoded secrets, instead instructing the user on how to authenticate securely through standard gcloud commands.

agent-platform-prompt-management