Observability-First Agent

The problem

AI agents often ship functional code that looks great in dev but acts as a black box in production. When requests fail, you are forced to grep unstructured logs across multiple services without any way to correlate events.

What it does

• Ensures every service ships with JSON structured logging instead of plain-text strings.
• Implements global correlation IDs to trace requests across microservice boundaries.
• Sets up standardized error boundaries that catch exceptions without leaking system internals.
• Builds comprehensive /health endpoints that monitor database, cache, and upstream status.
• Configures graceful shutdown handlers for SIGTERM to prevent data loss or hung connections.

Frameworks & tools

Designed for backend environments using Python (FastAPI, structlog), Node.js (Express, pino), and Go (slog). Compatible with OpenTelemetry and centralized logging stacks like Loki or CloudWatch.

Why this beats prompting it yourself

Standard prompts forget the "boring" 20 percent of production readiness like middleware for correlation IDs or redaction of PII. This skill enforces these patterns proactively, ensuring you don't have to refactor for observability after your first outage.

Use cases

• Spinning up new microservices that need to be production-ready on day one.
• Refactoring legacy handlers that lack proper error catching or request tracing.
• Standardizing telemetry across a polyglot architecture to simplify debugging.

Known limitations

Does not include automated dashboard creation or alerting rule definitions for specific providers.