smart-commit

Smart Commit — Atomic Commit Grouping

Analyze all uncommitted changes in the working directory, intelligently group related changes into atomic commits, and generate conventional commit messages for each group with clear motivation and context.

Step 1: Gather All Uncommitted Changes

# Get full status including staged, unstaged, and untracked files
git status --short

# Get diff of unstaged tracked changes
git diff

# Get diff of staged changes
git diff --cached

# Get list of untracked files
git ls-files --others --exclude-standard

Note: all four commands are needed to get a complete picture. Untracked files appear only in git ls-files --others. Staged changes appear only in git diff --cached.

Step 2: Analyze and Group Changes

For each changed file, analyze:

• Purpose: What is this change doing? (feature, fix, refactor, etc.)
• Scope: Which module, component, or feature does it affect?
• Relationship: Which other changes are logically related?

Group changes into atomic commits using these rules:

Grouping Rules

1. Single Responsibility: Each commit does ONE logical thing.
2. Feature Coherence: Changes to the same feature belong together.
3. Type Coherence: Do not mix feat with fix in the same commit unless tightly coupled.
4. File Relationships:
   • Component + its tests = same commit
   • Component + its styles = same commit
   • Config changes = separate commit (usually chore)
   • Documentation = separate commit (usually docs)
5. Dependency Order: If commit B depends on commit A, A comes first.

Type Reference

• feat: New functionality — group with related tests and types.
• fix: Bug fix — group with regression test if added.
• refactor: Code restructuring — can be larger, but still single purpose.
• chore: Deps, config, tooling — usually standalone.
• docs: Documentation only — standalone.
• style: Formatting — standalone, or omit if entangled with other changes.

Step 3: Present Commit Plan