Regex

Instructions

Handle the regex task in $ARGUMENTS. Two modes — detect which from the request:

Mode A — build a regex from a description

1. Clarify the target flavor if it matters: regex dialects differ (PCRE, JavaScript, Python re, Go RE2, POSIX, grep/ripgrep). Infer from context (the file type or tool mentioned); if genuinely ambiguous, ask or state the flavor you're assuming.
2. Write the pattern, then explain it token by token so the user understands and can adjust it.
3. Show test cases: a table of inputs that should match and inputs that should not, with the expected result. Cover the tricky edges (empty string, boundaries, unicode/accents, greedy vs. lazy).
4. Give a ready-to-paste snippet in the user's language showing how to use it (with the right flags — i, m, g, etc.).
5. Warn about footguns you introduced or avoided: catastrophic backtracking, unanchored matches, unescaped metacharacters, . matching newlines.

Mode B — explain or debug an existing regex

1. Break it into components and explain what each does in plain language.
2. State exactly what it matches and what it rejects, with examples.
3. If the user says it's not working, identify why (missing anchor, wrong flag, escaping, greedy quantifier) and give the corrected version.

Rules

1. Prefer readable, maintainable patterns over clever one-liners. Suggest named groups or verbose/x mode for complex patterns.
2. Always provide concrete match / no-match examples — never hand over a bare pattern.
3. Flag any catastrophic-backtracking risk and offer a safer rewrite.
4. If a regex is the wrong tool (e.g. parsing HTML or deeply nested structures), say so and suggest the proper approach.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Pradnyeshp
• Source: Pradnyeshp/Claude-Skills
• License: MIT

Install and usage instructions live in the source repository linked above.