What problem does it solve? Choosing between regex and LLM-based parsing for structured text is often done by guesswork, leading to silent data loss, unaccounted malformed records, and unjustified model costs. This Skill provides a decision framework and a complete parser pattern that accounts for every input record. ## Core Features & Use Cases - Decision Framework: A five-step process for choosing deterministic parsing versus model assistance based on grammar stability and measured accuracy against a labeled corpus. - Complete Parser Pattern: A Python reference implementation using dataclasses and regex that reports unmatched spans, duplicate IDs, and malformed records instead of dropping them. - Validator Boundary: A contract for optionally passing only problematic spans to an explicitly authorized model validator with schema checks, source-grounding verification, and cost caps. - Use Case: When converting a corpus of numbered quiz questions with A-D answers into structured records, use the deterministic parser first, then route only unresolved records to review or an approved validator while reporting exact-match accuracy and unresolved rates. ## Quick Start Ask the assistant to apply the regex versus LLM decision framework to parse your structured text file and report all unresolved records with reasons.
