/**
 * Plain text to the narrowest Atlassian Document Format that represents it.
 *
 * Jira's rich-text fields - a comment body, an issue description - are document
 * trees. Accepting one from a caller would mean accepting arbitrary structure
 * through a field that reads like "text": panels, mentions, embedded content,
 * links to anywhere. So JAM's contract is plain text in both places, and this
 * is the single conversion that produces the document.
 *
 * Shared rather than duplicated per field on purpose. Two copies of this would
 * be two answers to "what can an agent put in a Jira document", and the second
 * one would drift.
 *
 * Blank lines separate paragraphs; everything else is literal. No markdown is
 * interpreted, so text containing `*` or `#` says what it says.
 */
export declare function textToAdf(text: string): unknown;
/**
 * The text as it will exist once Jira has it.
 *
 * A write is only verified if what a direct read shows can be compared to what
 * was asked for - and for a rich-text field those two are never byte-identical.
 * The caller's string becomes a document, Jira stores the document, and reading
 * it back renders a document into text again. Blank-line runs collapse, block
 * edges lose their whitespace, a trailing newline disappears.
 *
 * None of that changes what the description says, so comparing raw strings
 * would fail every time. Comparing canonical forms fails only when the text
 * actually differs.
 *
 * Deliberately the same normalization `textToAdf` performs, from the same
 * place: "what JAM sends" and "what JAM will accept as proof it arrived" must
 * not be able to drift into two answers.
 */
export declare function canonicalizePlainText(text: string): string;