The Agent Skills Directory

Command Execution via gcloud CLI: The skill facilitates management of access controls through terminal commands. While this is the intended functionality for infrastructure-related tasks, it requires care to ensure commands are executed in the correct environment.
Usage Attribution Monitoring: The skill incorporates identifiers in command metadata for usage analytics. This involves data transmission to the service provider, but it is limited to skill-specific identifiers and provides a clear mechanism for users to opt out.
Safety Protocol for Policy Mutations: For all operations that change security state, a confirmation step is required. This is a critical guardrail that prevents the agent from making autonomous changes that could impact security posture without human oversight.
Prevention of Unsafe Access Configurations: Instructions are included to specifically block configurations that would create public access. This proactive check reduces the likelihood of accidental exposure, which is a significant risk in identity management.
Indirect Prompt Injection Surface Analysis: The skill processes user-defined configurations, which represents an ingestion point for potentially untrusted data. Although this creates an attack surface, the risk is mitigated by the mandatory confirmation boundary marker and the availability of verification tools, ensuring the agent's capability inventory is managed safely.

iam-helper-for-policy-management