build-observability

What problem does it solve?

Build observability for agentic systems is often missing at the exact moment operators need it most, leaving you with only raw logs and no reliable way to answer what is running, what changed, what artifacts exist, and how the run ended.

Core Features & Use Cases

• Normalized runs/events/commands contract: Projects long-running execution into stable, query-friendly tables for run status, timeline facts, and command lifecycles.
• Operator-focused dashboard surface: Exposes current stage, major event timeline, command activity, changed files, artifacts, sub-agent hierarchy, and deploy outcomes.
• Runtime collector projection pattern: Adapts local agent runtime artifacts (files, transcripts, logs, session metadata) into the canonical schema without hard-coding foreign layouts.
• Trace enrichment for memory/decision layers: Adds decision context so observability can become episodic inputs for memory and context-graph systems.
• Storage policy for local-first and shared access: Defaults to SQLite for local harnesses and uses Postgres when shared/remote readers are needed.

Quick Start

Use the build-observability skill to instrument your agent runtime by projecting its session/transcript artifacts into the normalized runs/events/commands schema and wiring the dashboard to those projections.