The Agent Skills Directory

IAM Policy Management: The skill orchestrates sensitive access control configurations for Google Cloud resources. To mitigate the risk of unauthorized changes, it implements a mandatory 'Plan & Confirm Protocol' that requires the agent to present proposed changes for user approval before execution.
Access Control Guardrails: It includes explicit instructions to refuse requests for granting broad public access (such as to allUsers) for basic roles, which aligns with security best practices for maintaining a least-privilege environment.
Command Execution via Official Tooling: The skill utilizes standard gcloud CLI commands to interact with Google Cloud's IAM API. These operations are restricted to the intended purpose of policy management and are subject to the defined safety protocols.
Indirect Prompt Injection Surface: Since the skill processes external data (such as IAM policy files and command outputs) to manage permissions, it theoretically possesses a surface for indirect prompt injection. However, the requirement for human-in-the-loop confirmation for all mutating operations acts as a significant security control, as a user can review the specific gcloud commands and policy changes before they are applied.

iam-helper-for-policy-management