grok

<!-- CAPABILITIES_SUMMARY: - regex_design: Safe regex authoring with anchors, lookaround, unicode flags - redos_prevention: Catastrophic backtracking detection, exponential complexity analysis - regex_engine_awareness: RE2 / PCRE / ECMAScript (ES2025 RegExp.escape, inline modifiers) / Oniguruma differences; Unicode 16.0 script property support by engine - parser_generator_selection: ANTLR4 vs PEG.js vs nearley vs tree-sitter vs chevrotain vs hand-written RD - parser_combinator_design: Parsec-style composable parsers, ts-parsec, chevrotain fluent API - grammar_ambiguity_detection: LALR conflicts, PEG ordered-choice hazards, left-recursion - internal_dsl_architecture: Fluent API, template-literal, s-expr, YAML-embedded, builder pattern - ast_design: Tagged union nodes, visitor pattern, immutable vs mutable trees - ast_transformation: Babel plugin, jscodeshift, ts-morph, tree-sitter query, JetBrains MPS - tokenizer_design: Lexer modes, context-sensitive tokens, indentation-based (Python-like) - error_recovery: Panic mode, phrase-level recovery, diagnostic quality (Elm-style) - grammar_evolution: Backward-compat rule additions, deprecation, version gates - lexer_design: Standalone tokenizer design (separation rationale, off-side rule, hand-written vs generator, lookahead, trivia) - error_design: Parser error-recovery + diagnostics (panic-mode, phrase-level, error productions, multi-span, expected-token reporting) - incremental_parsing: Incremental reparse (edit-aware state, dirty...