Ci Auto Fix

# CI Auto-Fix Command

Automatically fixes all CI failures for a given PR using intelligent analysis and iterative fixes.

## Usage

```bash
claude --command xats-dev:ci-auto-fix --pr <PR_NUMBER>

Description

This command performs a comprehensive CI fix workflow:

1. Checkout PR Branch - Fetches and checks out the PR branch
2. Analyzes CI Failures - Uses gh pr checks <PR> --watch to wait for CI completion
3. Diagnoses Root Causes - Parses error messages to understand the issues
4. Applies Fixes - Makes code changes to resolve identified problems
5. Tests Locally - Runs build, lint, and test commands to verify fixes
6. Commits and Pushes - Creates clear commits describing the fixes
7. Waits for CI Results - Uses gh pr checks <PR> --watch to monitor new CI run
8. Iterates - Repeats steps 2-7 until all CI checks pass (max 10 attempts)

Features

• Intelligent Error Analysis - Understands TypeScript, ESLint, test failures, and build errors
• Automated Fix Application - Applies appropriate fixes based on error types
• Local Verification - Tests changes locally before pushing
• Progress Tracking - Provides real-time updates on fix progress
• Retry Logic - Automatically retries if initial fixes don't resolve all issues
• Comprehensive Logging - Maintains detailed logs of all actions taken

Parameters

• --pr <number> - The PR number to fix (required)
• --max-retries <number> - Maximum fix attempts (default: 10)
• --wait-time <seconds> - Time to wait for CI to complete (default: 600)

Examples

# Fix CI for PR #135
claude --command xats-dev:ci-auto-fix --pr 135

# Fix with custom retry limit
claude --command xats-dev:ci-auto-fix --pr 135 --max-retries 5

# Fix with extended wait time
claude --command xats-dev:ci-auto-fix --pr 135 --wait-time 900

Error Types Handled

• TypeScript compilation errors
• ESLint violations
• Prettier formatting issues
• Module resolution failures
• Test failures
• Build configuration issues
• Missing dependencies
• Import order problems
• Type mismatches

Implementation

The command uses:

• debugger agent for error analysis
• test-engineer agent for test fixes
• code-auditor agent for code quality
• GitHub CLI for CI interaction
• Local test runners for verification

IMPORTANT EXECUTION INSTRUCTIONS

When executing this command, you MUST:

1. ALWAYS use gh pr checks <PR> --watch to wait for CI to complete before analyzing results
2. NEVER declare success without verifying all CI checks have passed
3. ITERATE until all checks pass or max retries reached:
   • After pushing fixes, use gh pr checks <PR> --watch again
   • Analyze any remaining failures
   • Apply additional fixes if needed
   • Repeat until all checks are green
4. Track iteration count and stop after max retries (default: 10)
5. Report final status with:
   • Number of iterations performed
   • Final CI status (all passing or list of remaining failures)
   • Summary of all fixes applied

Workflow Loop

for iteration in 1..max_retries:
  1. Run: gh pr checks <PR> --watch
  2. If all checks pass: SUCCESS - exit loop
  3. If checks fail: analyze failures
  4. Apply fixes based on error types
  5. Test locally (pnpm lint, pnpm test, pnpm build)
  6. Commit and push fixes
  7. Continue to next iteration

If max_retries reached without success, report remaining failures. ```