The Agent Skills Directory

Command Execution Considerations: The skill utilizes gcloud CLI commands to interact with Google Cloud Security Command Center. It includes instructions to configure regional API endpoints via gcloud config set, which modifies the local CLI environment to ensure data residency compliance when querying findings. These commands are vendor-specific and necessary for the intended functionality.
Indirect Prompt Injection Surface: The skill processes security finding data (JSON payloads) retrieved from Google Cloud Security Command Center.
Ingestion Points: Data enters the agent's context through the output of gcloud scc findings list and gcloud scc findings group commands.
Boundary Markers: The skill includes a 'Handoff' section that explicitly instructs the agent to pass retrieved payloads to specialized remediation or analyzer skills rather than executing changes itself, which helps maintain clear execution boundaries.
Capability Inventory: The available tools are restricted to read-only gcloud queries and localized CLI configuration. The skill prohibits the agent from querying underlying cloud resources like virtual machines, storage buckets, or IAM policies directly.
Sanitization Patterns: The instructions provide a clear schema and guidance for extracting specific fields (such as CVE IDs, package versions, and CVSS scores) from the finding payload, encouraging structured data handling and reducing the impact of raw external content.
Operational Scoping and Safety: The core execution rules enforce strict resource scoping by requiring explicit parent resource paths (organizations, projects, or folders). It also provides clear guidance to halt execution immediately if errors occur, if credentials expire, or if findings are ambiguous, which prevents unauthorized discovery attempts or runaway loops.

google-cloud-scc-query