article --summary
The summary
The post explains why database incident triage depends on tacit knowledge held by senior reliability engineers, and how encoding that expertise as a procedural Skill can make it available to an automated agent. It describes a single Skill covering five Aurora PostgreSQL failure modes: connection pool exhaustion, replication lag, slow query regression, storage pressure, and parameter drift. Each mode is reached through a signal classification table and a priority order that checks connection exhaustion first, since it can mask other symptoms.
The author walks through the Skill's directory layout, its SKILL.md frontmatter and decision logic, and the reference files that hold thresholds, query templates, and tuning guidance. The reader learns what prerequisites a deployment needs, including CloudWatch metrics, Performance Insights, and an Agent Space with access to the cluster's account.
The headline result is a mean time to resolution falling from 52 minutes to 14 minutes, a 73 percent reduction, measured across three simulated incidents in a controlled internal test. The sample is small, so the figure is best read as an indicative result rather than a broad benchmark.
Editorial note: This is a summary of an article first published by AWS DevOps Blog. The full piece is on their site.