Release v0.5.0

Fixed

• Fixed test categorization discrepancy in test runner
  • Unit tests with "System" in filename were incorrectly categorized as System tests
  • Test discovery showed System=8 but only 6 were actual System tests (2 were Unit tests)
  • Updated categorization logic to check folder path before filename pattern
  • Now correctly excludes Unit folder files from System test categorization
  • Consistent categorization priority: Tags > Folder Path > Filename Pattern
• Fixed git diff syntax error in publish script
  • Changed from incorrect origin/branch...HEAD to correct origin/branch..HEAD
  • Prevented script from continuing with broken state when commands fail

Added

• Comprehensive error handling for ALL commands in publish script
  • Every git command now checks $LASTEXITCODE
  • Every gh (GitHub CLI) command has error handling
  • File operations (Get-Content) use -ErrorAction Stop
  • Script exits immediately on any command failure
  • Clear error messages indicate exactly what failed
  • Prevents silent failures and broken release states

Changed

• Improved automated commit squashing in publish script
  • Removed manual prompt for squashing - now automatically combines commits
  • Integrated Claude CLI for automatic changelog verification
  • Uses claude -p command with piped input for seamless integration
  • Falls back gracefully if Claude CLI is not available
  • Eliminates manual copy-paste step for AI verification

Added

• Comprehensive changelog generation with full git diff analysis
  • Claude now receives complete git diff from origin/master to HEAD
  • Analyzes actual code changes, not just commit messages
  • Ensures 100% coverage of all changes in the changelog
  • Generates complete Unreleased section from scratch
  • Captures changes that might be missed in commit messages
  • No size limits - sends full diff for complete analysis
  • Proper error handling with exit on Claude failures
  • Saves prompt to file on error for manual processing