Testing distributed systems with AI agents

AI coding agents can now test distributed systems using two skills that generate claim-driven test plans and detailed findings reports with explicit blame classification, enabling reviewers to make shipping decisions without re-running tests. The approach enforces best practices from the field—starting from product claims rather than test cases, pairing chaos injection with abstract models and checkers, and providing 10-state verdicts that distinguish between actual failures and false passes. The workflow produces structured Markdown artifacts documenting architectural assumptions, coverage adequacy, and honest confidence statements about what remains unverified.