shaoxyz/git-commit-analyzer

Overview

This skill analyzes git commits with AI to judge whether your workday was productive or not using a playful "牛马" scale. It supports GitHub, GitLab, and local repositories and produces both pragmatic and blunt stylistic outputs (two voices: "Worker" and "Linus"). The goal is to measure meaningful contribution beyond raw lines and commit counts and to surface actionable signals and a daily vibe report.

How this skill works

The tool fetches commits for a time window or repo, computes machine-derived metrics (substance_score, bullshit_score, complexity, impact) and estimates a rewrite_index and business_value for each change. It then applies a scoring formula to produce final scores, badges, leaderboards, and two-styled narrative summaries that highlight irreplaceable work or possible fluff. Outputs are JSON reports and optional human-readable roasts and advice.

When to use it

• Daily or weekly team retros to add lighthearted insight into contribution quality
• Prepping a standup or demo to summarize meaningful changes and blockers
• Spotting trends like repeated low-value commits or sudden high-impact fixes
• Generating conversation starters for mentoring or code-review follow-ups
• Assessing where automation or context capture could reduce repetitive work

Best practices

• Treat reports as entertainment and diagnostics, not as performance reviews
• Include context (PR descriptions, issue links) to improve business-value scoring
• Run on a consistent cadence and compare trends instead of single-day snapshots
• Use the two voice outputs to surface both supportive feedback and blunt technical critique
• Combine AI signals with human judgment for mentoring or process changes

Example use cases

• Run before a daily to highlight who solved high-impact bugs vs. who did low-value refactors
• Team retrospective: identify contributors with high AI-survivor scores for mentorship pairing
• Manager check-in to spot repeating low-substance patterns and provide coaching
• Automated badge generation for weekly team shout-outs (e.g., "Online Firefighter")
• Code triage: surface commits likely to require human review because AI couldn't rewrite them