Scaling out Distroless adoption With AI

This post details the evolution of Grab's Distroless image adoption strategy by introducing an AI-powered agent to automate the scaffolding of medium tests for services that previously lacked them. The agent uses a multi-phase workflow (Detect, Scaffold, Fix, Raise MR, Monitor CI, Repeat) guided by skills and integrated with tools like Claude Code, Glean, and Sourcegraph via MCP. It focuses on test files and CI configurations, ensuring guardrails and human review before merging. The AI also assists in the 'Patch-Test-Compare' loop by identifying OS-level package dependencies, generating multi-stage builds, and verifying changes with existing medium tests.