Entity: runlog

Brief: Local, opt-out run log. Wraps every skill script's entry point to record its arguments, exit code, duration and capped stdout/stderr in ~/.ouro/runs/<project>-<hash>/events.jsonl, and lets agents log reports with a wiki snapshot. Nothing is transmitted.

Overview

ouro/scripts/runlog.py gives the maintainer ground truth about real-world runs, especially agent-driven onboarding in other projects. See ADR-012.

Integration

Each script ends with:

@snippet runlog-main

if __name__ == '__main__':
    import runlog  # sibling module; the script's directory is on sys.path
    runlog.run('capture', main)

This applies to bootstrap.py (whose inline __main__ code became main()), capture.py, claim_scan.py and hooks.py. Importing these modules (for example bootstrap.py importing hooks) doesn't log; only direct script runs do.

Functions

CLI

python <path-to-skill>/scripts/runlog.py report [--step onboarding] [--message TEXT]   # else reads stdin
python <path-to-skill>/scripts/runlog.py show [project_path] [--last 50]
python <path-to-skill>/scripts/runlog.py list

Event shapes

Note: The first real finding from the log came from a throwaway demo. After onboarding, the first capture.py --pop hands the agent CLAUDE.md, which is mostly the appended protocol boilerplate, because --crawl stages instruction files. Left as is until external test runs confirm how much it matters.
Warning: Logs contain repo-derived text (capture content, comment text, paths) in the user's home directory. OURO_RUNLOG=off disables logging; OURO_HOME relocates it.