SuperAPI AI Directory

auto-commit

Coding & development

View repository

Updated 10-15 22:17

About this skill

Auto-Commit Skill

🎯 Use Case Demo

Scenario

Developer Sarah just finished implementing a new seat selection feature. She has 5 files changed but struggles to write a good commit message.

Before (Manual):

git commit -m "updated stuff"  # 😬 Bad commit message

After (With Skill):

# Skill analyzes the diff and suggests:
feat(booking): add passenger seat selection with availability check
  • Add SeatSelection model with row and seat properties
  • Implement seat availability validation in BookingService
  • Add GET /api/seats/{flightId} endpoint
  • Update booking flow to include seat assignment

### Time Saved: 5-10 minutes per commit × 20 commits/day = 2+ hours/day

---

## When to Use This Skill

- After making changes, before committing
- When unsure how to describe changes
- To maintain consistent commit message format
- For generating detailed commit bodies

## Commit Message Format

Following [Conventional Commits](https://www.conventionalcommits.org/):

():

[optional body]

[optional footer]


**Types:**
- `feat` - New feature
- `fix` - Bug fix
- `docs` - Documentation
- `style` - Formatting
- `refactor` - Code restructuring
- `test` - Adding tests
- `chore` - Maintenance

## Agent Instructions

When user asks to commit or wants a commit message:

1. **Get staged changes**: `git diff --cached`
2. **Analyze the diff** to understand:
   - What files changed
   - What type of change (feature, fix, refactor)
   - The scope (which module/component)
3. **Generate commit message** following conventional commits
4. **Present to user** for approval
5. **Execute commit** if approved

### Example Prompts

User: "Commit my changes"
→ Analyze diff, suggest message, commit

User: "What should my commit message be?"
→ Analyze diff, suggest message only

User: "Commit with message: fix login bug"
→ Use provided message directly

## Demo Script

```bash
# 1. Make some changes
echo "// New feature" >> src/feature.ts

# 2. Stage changes
git add .

# 3. Ask agent: "Generate a commit message for my changes"

# 4. Agent analyzes and suggests:
#    feat(core): add new feature module
#    
#    - Add feature.ts with core functionality
#    - Implement base feature logic

# 5. User approves, agent commits

Benefits

Metric Before After Improvement
Commit message quality Inconsistent Standardized ✅ 100%
Time per commit 2-5 min 10 sec ⬇️ 90%
Changelog generation Manual Automated ✅ Enabled
Code review context Poor Excellent ⬆️ 10x