What problem does it solve? Manually auditing Google Cloud Storage buckets and projects for security misconfigurations is slow and error-prone, and isolated alerts miss dangerous combinations of individually low-risk settings. This Skill gathers real telemetry from your GCP project, evaluates baseline controls, and identifies toxic combinations of vulnerabilities mapped to Google's Secure AI Framework (SAIF). ## Core Features & Use Cases - Telemetry-Grounded Assessment: Runs preflight permission checks, then collects bucket- and object-level signals from Storage Insights BigQuery datasets plus project-level posture (org policies, VPC-SC, audit logs, Model Armor). - Toxic Combination Detection: Correlates multiple signals to flag critical exposures, such as a public bucket holding ML training data with no CMEK, no VPC-SC, and no audit logging. - SAIF-Aligned Reporting: Classifies bucket sensitivity (SDP, tags, or inferred), modulates severity, and produces a structured remediation report referencing SAIF risk categories. - Use Case: Ask the agent to assess a GCP project before deploying an AI workload; it verifies credentials, discovers Storage Insights datasets, evaluates UBLA/TLS/HMAC/audit-log controls, and outputs prioritized findings with fixes. ## Quick Start Ask the agent to run a security assessment on your GCP project by providing the project ID, for example: assess the security posture of project my-gcp-project.
