log-message-improver

Purpose

Improve logging statements so they're actually useful during an incident or investigation — the right level, structured (not just string concatenation), and carrying the context a future reader will need without access to the original author's memory of what was happening.

When to Use

• Logs are too sparse to debug an issue from, or too noisy to find the signal.
• Adding logging to new code and want it done well from the start.
• Reviewing logging statements in a PR.

What to Analyze

1. Log level appropriateness — is this actually an error (something broke) or just unusual-but-expected (should be warn/info)? Overuse of error for expected conditions trains people to ignore error-level alerts.
2. Structured vs. string-concatenated — log.info(\User ${id} did ${action}`)is harder to query/filter thanlog.info('user_action', { userId: id, action })` — prefer structured fields where the logging framework supports it.
3. Missing context — does the log include enough identifying information (request ID, user ID, relevant entity IDs) to correlate it with other logs from the same request/flow during an investigation?
4. Missing "why," not just "what" — log.warn('retrying') is less useful than log.warn('retrying', { attempt: 2, reason: 'timeout', nextRetryMs: 500 }).
5. Noise — logging inside a hot loop at info/debug level that will flood logs without adding investigative value; logging the same event redundantly at multiple layers.
6. Sensitive data in logs — flag any logging of passwords, tokens, full credit card numbers, or other sensitive data that shouldn't be persisted in log storage.

Output Format

• Each finding: the log statement, what's wrong (level/structure/context/noise/sensitive-data), and the improved version as actual code.
• Group by issue type if reviewing many at once (Level Issues, Missing Context, Noise, Sensitive Data).
• Flag sensitive-data logging as highest priority regardless of other issues.

Avoid

• Recommending logging be added everywhere indiscriminately — more logs isn't the goal, useful logs are; flag genuinely under-logged critical paths, not every line.
• Assuming a specific logging framework's API without checking what's actually in use in the codebase.
• Suggesting structured logging changes that aren't supported by the project's actual logging setup — check what's available first.