| name | logstrip |
| description | Compress noisy CI, build, runtime, scanner, and test logs before Codex analyzes them. Use when the user asks to inspect, trim, reduce, summarize, or prepare logs for diagnosis, or when the user pastes log-like output (timestamps, log levels, stack traces, CI markers). |
LogStrip Log Compression
LogStrip removes repeated low-signal lines, masks common identifiers, collapses internal stack frames, and reports token savings so Codex spends context on diagnostics instead of noise.
Procedure
- Identify the raw input log and desired output path. If no output path is supplied, use
<input>.logstrip.log.
- Verify
logstrip is installed. If not, instruct the user to run npm i -g logstrip.
- Run
logstrip <input> -o <output> --stats, adding --aggressiveness <level> only when requested.
- Diagnose from the compressed output file.
- Report the output path, savings percentage, input/output token estimates, dropped lines, duplicate lines, and hidden internal stack lines.
Guardrails
- Do not print full raw logs unless the user explicitly requests excerpts.
- Do not install packages automatically.
- Do not use
--json unless an output path is present.
- Keep stats separate from compressed content when piping.
