What problem does it solve?
AI-assisted coding sessions leave no trace of who contributed what or why decisions were made. This Skill analyzes your Claude Code session and git history to produce calibrated evidence of AI versus developer contributions, decisions, and reasoning.
Core Features & Use Cases
- Signal Detection: Classifies sessions as HIGH, MEDIUM, or LOW signal using git diff stats and conversation analysis, so routine sessions produce no noise.
- Four-Frame Intent Classification: Scores technical complexity, uncertainty, decision forks, and AI contribution to classify sessions as FEATURE_BUILDING, BUG_FIXING, REFACTORING, and more.
- Calibrated Evidence Artifacts: Generates DECISIONS.md entries, session-history markdown, WORKLOG lines, and a self-contained HTML proof file with token usage stats.
- Use Case: After a debugging session where Claude identified a root cause, run the skill to append a structured bug-fix record to DECISIONS.md and generate a shareable HTML proof for your portfolio.
Quick Start
Ask Claude to run /collab-proof at the end of your coding session to analyze what was built and generate the collaboration evidence artifacts.