logging-sucks

This skill is adapted from "Logging sucks. And here's how to make it better." by Boris Tane.

When helping with logging, observability, or debugging strategies, follow these principles:

Core Philosophy

• Logs are optimized for querying, not writing — always design with debugging in mind
• Context is everything — a log without correlation IDs is useless in distributed systems
• Logs are for humans during incidents, not just for compliance or "just in case"
• If you can't filter and search your logs effectively, they provide zero value
• Mental model shift: Log what happened to this request, not what your code is doing

Wide Events / Canonical Log Lines

Instead of scattering 10-20 log lines throughout a request, emit one comprehensive event per request per service. This is the most important concept for effective logging.

• Build the event object throughout the request lifecycle
• Enrich it with context as you process (user info, business data, feature flags)
• Emit once at the end with all context attached
• Include 30-50+ fields containing everything useful for debugging

Example wide event structure:

{