Project Instructions

Commands

• One file per command or subcommand. Keep the entry point thin.
• Use a proper argument parser (commander, yargs, argparse, click). Don’t parse argv manually.
• Every command should have —help text that’s actually helpful.

Output

• Print results to stdout. Print errors and diagnostics to stderr.
• Support —json or —quiet flags for scripted usage.
• Use exit codes correctly: 0 for success, 1 for errors, 2 for usage mistakes.

Error Handling

• Show clear error messages. Include what went wrong and what to do about it.
• Don’t print stack traces to users unless —verbose is set.
• Fail fast on bad input. Don’t continue with partial arguments.

Testing

• Test commands end-to-end by capturing stdout/stderr and checking exit codes.
• Test edge cases: missing arguments, invalid input, empty files.

Git

• Write short commit messages. One feature or fix per commit.