Log error summarizer: group errors by masked signature

What it does

Instead of reading a multi-megabyte log into context, get a short ranked list of which errors happen, how often, and when.

• Normalizes messages by masking numbers, UUIDs, IPs, e-mails, URLs and paths, so the same error with different values becomes one signature
• Shows count, share, first/last seen, files, exception line, most common stack frames and raw sample lines
• Attaches multi-line stack traces (Python, Java, Node.js, Go, .NET, Ruby) to the record they belong to
• Plain-text logs and JSON Lines (level / msg / stack keys auto-detected), .gz files, multiple files and stdin
• Markdown or JSON output

The agent then explains likely causes in order of impact and proposes next checks.

How to use

python scripts/log_summary.py app.log
python scripts/log_summary.py "logs/*.log*" --levels WARN,ERROR,FATAL --top 20
python scripts/log_summary.py app.jsonl --json --out summary.json

Ask Claude, Codex, Hermes Agent or Cursor to "summarize the errors in this log" and it follows the skill's procedure.

Requirements

• Python 3.10+ (standard library only)

Caveats

• Read-only: logs are never modified or deleted, and nothing is sent over the network.
• Only signatures are masked; sample lines are raw, so take care with personal data and secrets when sharing.
• With unusual custom formats a stack trace may be counted as separate records.