Field Guide 05: a three-level readiness test for agent-safe design systems. Level 1 (readability) asks whether an agent finds the right component from the docs alone, Level 2 (executability) whether generated UI is checked mechanically, Level 3 (survivability) whether internals can change without breaking consumers. The kit is a zero-dependency Node scanner that reads a repository and reports the level reached, with a clean example repository that reaches Level 3, a drifting one that stops at Level 1 with 10 failures, and a self-assessment template for teams without one repository to scan.
Before you point agents or copilots at your design system, when generated UI keeps inventing props or raw values, or as a quarterly health check with a minimum level enforced in CI.