evolution-scan

Run the evolution_timeline.py script against the repository's git-history.txt and git-commits.txt exports to get a JSON digest of periods, areas, people, and lifecycle events. The scanner then narrates that digest into era, growth, focus-shift, people, lifecycle, and trajectory findings.

It requires a completed Sokrates analysis with git history, specifically git-history.txt and git-commits.txt at the source root, or zips/git-history.zip inside the extracted data.zip. Supplying the extracted data directory additionally enables current file inventory, deletion detection, and area size figures.

The timeline script exits with code 3 and the scan writes a single info finding in the trajectory group stating that evolution cannot be measured, with stats.history_data set to absent. It never reconstructs history from file dates or guesswork.

The script classifies each author as active, fading, or gone based on their last commit date relative to a threshold of 180 days or a quarter of the history. It also lists arrivals in the last year, fading major contributors whose 90-day output halved, and rising contributors.

Only when commit messages are classifiable. The digest reports theme_other_pct, and when over 30 percent of messages are unclassified the theme trend is flagged unreliable, so the work mix should be judged from notable commits instead of keyword themes.

Shares count the percentage of a period's commits that touch each area, and a commit touching three areas counts in all three. They should be read as 'touched by 45 percent of commits', never as a partition of the work.