Evidence over Anecdotes: Running A/B Tests on AI Agent Tooling
PagerDuty's engineering team found that splitting its engineering standards into domain-specific AI agent skills did not improve quality, with the two designs tying when the right skill loaded, but introduced a silent fa…