Reserved characters are interpreted as syntax only when they form acomplete construct. Once a reserved prefix appears, the parser commits toparsing that construct and errors if it is not well-formed — it does notsilently fall back to treating the input as plain text. Malformed use of areserved prefix is a hard error, not a fallback to text.
*, _, ` , $, \ are formatting markers. A bare or unclosedoccurrence is a parse error: in formatted text such a character should bewritten as inline code or escaped, not exposed as special-looking plain text.Escape sequences (\*, \_, \` , \$, \\) produce a literalcharacter; escaping any other character is an error.
The []() link syntax is not supported; [text] and [text](url) are kept asplain text. [ and ] are only special as the footnote reference prefix[^; a bare [ or ] is ordinary text. ! is ordinary punctuation, so anexclamation mark in prose is parsed as plain text.
The grammar avoids backtracking: every construct starts with either a fixedatomic prefix or a character unique to that construct. Alternatives havedisjoint first tokens, so each input has a single parse and there is noexponential blow-up on adversarial input such as runs of unclosed brackets.