a1k7/governance-score — reverse-engineered prompt

Build me a simple Python command line tool that scores how governable an AI agent run was from a local JSON trace file.

The user should be able to run something like python governance_score.py trace.json and get a clear report with scores from 0 to 100 for continuity, evidence freshness, rollback viability, and one overall governance score. The input is a JSON array of steps where each step says whether continuity was valid, evidence was fresh, rollback was viable, and whether there was a hidden commitment.

Please include a readable terminal report, a verbose mode that explains each step, and a JSON output mode for automation. Keep it offline, with no API or cloud calls, and only use the Python standard library. Add a few example trace files and basic tests so I can quickly verify the scoring formula and interpretations like Excellent, Good, Warning, and Critical.