Writes regex patterns from natural-language descriptions, with three match + three no-match test cases (including a near-miss) and at least one edge case. Identifies the right flavor (JS / Python / PCRE / POSIX BRE-vs-ERE / RE2) for the user's environment, flags catastrophic backtracking, and recommends parsers / libraries when regex is the wrong tool (HTML, JSON, email, URL). Example: "what's the regex for ISO 8601 dates?"