agent-bom discover gcp

This Skill addresses the challenge of systematically inventorying AI Agent and MCP-related assets in GCP environments. In practice, teams often rely on manual logs or scattered gcloud commands to catalog resources like Vertex AI, Cloud Run, Cloud Functions, and GKE, which is prone to omissions and fails to produce a unified, exchangeable inventory. This tool automatically discovers these assets and outputs a standardized agent-bom JSON, providing a structured data foundation for subsequent scanning, auditing, or migration.

Usage is straightforward: simply install the agent-bom library in an environment configured with GCP Application Default Credentials (ADC) or Workload Identity. Then, run a single command specifying the project ID and region to trigger the auto-discovery process. The tool calls GCP read-only APIs to collect resource information and writes the results to your specified JSON file. The entire process requires no long-term credentials, ensuring high security.

It is well-suited for teams or individuals needing unified management of AI infrastructure on GCP, especially those who have deployed multiple Vertex AI models, Cloud Run services, or GKE clusters but lack a global asset view. For scenarios requiring regular generation of asset inventories for compliance audits or cost analysis, this Skill significantly boosts efficiency.

We recommend integrating this tool into CI/CD pipelines or regular inspection tasks to automatically generate asset snapshots and track changes. Note that it primarily relies on GCP read-only permissions; ensure the service account or user has the necessary permissions like resourcemanager.projects.get, aiplatform.models.list, etc. Also, securely store the output JSON file to avoid exposing project metadata.