Smart Commit
Trigger
Use when saying "commit", "save changes", or ready to commit after making changes.
Workflow
- Check current state and identify what to commit.
- Run quality gates (lint, typecheck, tests on affected files).
- Scan staged changes for issues.
- Draft a conventional commit message from the diff.
- Stage specific files, create the commit.
- Prompt for learnings from this change.
Commands
git status
git diff --stat
npm run lint 2>&1 | tail -5
npm run typecheck 2>&1 | tail -5
npm test -- --changed --passWithNoTests 2>&1 | tail -10
git add <specific files>
git commit -m "<type>(<scope>): <summary>"Code Review Scan
Before committing, check staged changes in production code (not test files) for:
console.log/debuggerstatements (suppressed in test files — see Review Suppressions)- TODO/FIXME/HACK comments without ticket references (e.g.,
TODO(JIRA-123)is fine) - Hardcoded secrets or API keys
- Leftover test-only code
Flag any issues before proceeding.
Commit Message Format
<type>(<scope>): <short summary>
<body - what changed and why>Types: feat, fix, refactor, test, docs, chore, perf, ci, style
Guardrails
- Never skip quality gates unless user explicitly says to.
- Stage specific files by name. Never
git add -Aorgit add.. - Summary under 72 characters. Body explains *why*, not *what*.
- No generic messages ("fix bug", "update code").
- Reference issue numbers when applicable.
Output
- Quality gate results (pass/fail)
- Issues found in staged changes
- Suggested commit message
- Commit hash after committing
- Prompt: any learnings to capture?
Review Suppressions
Do NOT flag these during the pre-commit scan. They add noise without catching real bugs:
- Threshold, config value, or feature flag changes (limits, timeouts, retry counts)
- Import reordering that does not change runtime behavior
- Whitespace-only or formatting-only changes
- Adding or removing
console.login test files - TODO/FIXME comments (tracked separately in issue trackers)
- Variable or parameter renames that do not change behavior