commit-push

Commit Push (Fully Autonomous)

Overview

This skill provides a fully autonomous pre-commit workflow that requires ZERO user interaction. It verifies code quality, fixes all issues automatically, commits with a meaningful message, and pushes to remote. Uses parallel agents for speed.

CRITICAL: No Permission Required

This workflow is PRE-AUTHORIZED to:

• Run all verification commands (lint, tsc, test, build)
• Automatically fix any issues found
• Stage all changes with git add .
• Commit without asking for confirmation
• Push to remote without asking for confirmation

DO NOT ask the user for:

• Permission to fix issues
• Confirmation of commit message
• Approval to push
• Any other confirmation

Just execute the entire workflow autonomously and report the results at the end.

When to Use This Skill

• User explicitly requests to commit and push changes
• User asks to "verify and push" or "check and commit"
• After completing a feature or bug fix that needs to be committed

Autonomous Workflow

Phase 1: Parallel Verification

Run ALL THREE verification steps in parallel using Task agents in a SINGLE message.

Launch three agents simultaneously with ONE message containing THREE Task tool calls:

Task 1: Linting Agent (subagent_type: Bash)
Task 2: Type Checking Agent (subagent_type: Bash)
Task 3: Test Agent (subagent_type: Bash)

Agent 1: ESLint Verification

Prompt:

cd /Users/ohadfisher/git/boggle-new/fe-next
Run: npm run lint
If issues found, run: npm run lint -- --fix
For remaining issues: read files, apply manual fixes per CLAUDE.md guidelines.
Re-run lint to verify zero errors.
Report: linting status and any fixes applied.
Do NOT ask for user permission - fix everything automatically.

Agent 2: Type Checking Verification

Prompt:

cd /Users/ohadfisher/git/boggle-new/fe-next
Run: npx tsc --noEmit
If errors found: read files, fix type errors. Never use 'any' type.
Re-run to verify zero errors.
Report: type checking status and any fixes applied.
Do NOT ask for user permission - fix everything automatically.

Agent 3: Test Verification

Prompt:

cd /Users/ohadfisher/git/boggle-new/fe-next
Run: npm run test
If failures: analyze each test failure, fix root cause in code (not tests).
Re-run to verify all tests pass.
Report: test status and any fixes applied.
Do NOT ask for user permission - fix everything automatically.

Wait for all three agents to complete.

Phase 2: Build Verification

After Phase 1 completes successfully:

cd /Users/ohadfisher/git/boggle-new/fe-next && npm run build

If build fails:

• Fix compilation/bundling issues immediately
• Re-run until build succeeds
• DO NOT ask for permission - just fix it

Phase 3: Commit and Push (Automatic)

After build succeeds, execute these commands in sequence:

Step 1: Check Current State

cd /Users/ohadfisher/git/boggle-new/fe-next && git status && git diff --stat

Step 2: Generate Commit Message

Analyze the changes and create a conventional commit message:

• Use: feat:, fix:, refactor:, test:, docs:, chore:
• First line: Brief summary (50-72 characters)
• Body: Detailed explanation if needed
• Always include: Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Step 3: Stage and Commit

cd /Users/ohadfisher/git/boggle-new/fe-next && git add . && git commit -m "$(cat <<'EOF'
[Your generated commit message]

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
EOF
)"

DO NOT ask for confirmation. Just commit.

Step 4: Push to Remote

cd /Users/ohadfisher/git/boggle-new/fe-next && git push

If push fails:

# Set upstream if needed
git push -u origin $(git rev-parse --abbrev-ref HEAD)

If still fails due to diverged branches:

git pull --rebase && git push

DO NOT ask for confirmation. Just push.

Parallel Execution Template

When executing Phase 1, use this exact pattern:

[Send ONE message with THREE Task tool calls]

Task 1:
  subagent_type: Bash
  description: "Run linting"
  prompt: [Linting agent prompt]

Task 2:
  subagent_type: Bash
  description: "Check types"
  prompt: [Type checking agent prompt]

Task 3:
  subagent_type: Bash
  description: "Run tests"
  prompt: [Test agent prompt]

Error Handling (Autonomous)

• Lint errors: Run --fix, then manual fixes
• Type errors: Fix types properly (no any)
• Test failures: Fix the code, not the tests
• Build failures: Fix compilation issues
• Push failures: Handle upstream/rebase automatically

NEVER stop to ask the user. Fix and continue.

Success Criteria

• ✅ ESLint passes (zero errors)
• ✅ TypeScript passes (zero errors)
• ✅ All tests pass
• ✅ Build succeeds
• ✅ Changes committed
• ✅ Changes pushed

Final Report

After ALL steps complete, provide this summary:

✅ Commit and Push Complete

Verification Results:
- Linting: [Status] [Fixed X issues if any]
- Type checking: [Status] [Fixed X errors if any]
- Tests: [Status] [Fixed X failures if any]
- Build: [Passed]

Git:
- Commit: [hash]
- Message: [first line]
- Branch: [branch name]
- Remote: [pushed successfully]

Important Notes

1. This is a FIRE-AND-FORGET skill - user invokes it and waits for completion
2. All operations are pre-authorized - no confirmation dialogs
3. Fix everything automatically - don't ask, just fix
4. Report at the end - only communicate when complete
5. Use parallel agents - maximize speed by running verification in parallel