change-classification

What problem does it solve?

Prevents accidental, out-of-context, or unsafe changes from being committed by classifying every changed file as in-context, uncertain, or out-of-context and flagging secrets, large additions, and red-flag patterns before staging or committing.

Core Features & Use Cases

• Signal-based classification: Uses primary signals from branch diffs and issue/task context and secondary signals like sibling changes and test associations to determine relevance.
• Safety and red flags: Detects secret patterns, large binaries, lockfile changes, and auto-generated files and enforces blocking or warning actions for critical patterns.
• Review focus and grouping: Identifies first-touch files and groups in-context changes into atomic commits by logical unit, type, or directory to improve review quality.
• Use Case: During a feature branch commit, run classification to automatically surface files that are unrelated to the issue, highlight first-touch additions for extra review, and produce atomic commit suggestions.

Quick Start

Classify all changed files and display a table showing file-level classifications and signals before staging or committing.