reliability-ops

Host health monitoring identifies anomalies and capacity pressure across CPU, memory, disk, network, and GPU metrics. This skill captures live service degradation and persists findings into a SQLite-backed runtime for durable reliability context.

Recurring reliability reviews apply local observability scripts to gather health signals and document service failures. Results persist in a shared SQLite-backed kernel, ensuring continuity and verification state across ongoing assessments.

Capacity planning and GPU data correlation are supported natively. The skill captures and correlates logs, endpoints, and GPU metrics to detect capacity pressure and support post-incident reviews for slow or failing services.

A SQLite-backed kernel transforms observability findings into durable knowledge by recording tickets, verification state, and continuity. Standalone notes do not count as durable knowledge because they lack persistent runtime state for ongoing reliability operations.

Local helper scripts gather and persist health metrics to bootstrap health graphs. This provides concrete observability by turning raw system signals into structured health assessments stored within the shared reliability operations kernel.