Build and Explain a Regular Expression

This prompt gets you a working regular expression plus the thing most regex tools skip: a plain-English breakdown of every piece and a built-in test suite, so you can verify it does what you think before shipping it. It's built for the moment you know exactly what you want to match but don't want to reverse-engineer character classes and lookaheads by hand. Specifying the flavor matters because regex syntax differs across JavaScript, Python, and PCRE — the same pattern can behave differently or error out.

What you can expect back

Pattern:
/^\(?\d{3}\)?[-.\s]?\d{3}[-.\s]?\d{4}$/

Breakdown:
- ^                start of string
- \(?              optional opening parenthesis
- \d{3}            three-digit area code
- \)?              optional closing parenthesis
- [-.\s]?          optional separator: dash, dot, or space
- \d{3}            three-digit prefix
- [-.\s]?          optional separator
- \d{4}            four-digit line number
- $                end of string

Function:
function isUSPhone(input) {
  return /^\(?\d{3}\)?[-.\s]?\d{3}[-.\s]?\d{4}$/.test(input);
}

Tests:
- '(555) 123-4567' -> true
- '555-123-4567'   -> true
- '5551234567'     -> true
- '55-123-4567'    -> false (area code too short)
- '555-123-456'    -> false (line number too short)

Illustrative example — your results will vary by tool and inputs.

Get sharper results

• 01Always include the hardest edge cases in your description ('must reject numbers with letters, must allow an optional +1 country code') so the model designs for them instead of you discovering gaps later.
• 02Ask it to anchor with ^ and $ for full-string validation, or to drop them if you're searching for matches inside larger text — these serve opposite purposes.
• 03Have it add 'and explain the time complexity / any catastrophic backtracking risk' for patterns with nested quantifiers, since some regexes can hang on malicious input.
• 04Request the pattern in two flavors at once ('give me both the JavaScript and Python versions') when your validation runs on both front end and back end.
• 05Paste your real failing inputs back to the model and ask it to add them as test cases, then regenerate — this catches mistakes faster than eyeballing the pattern.

Adapt it for your case