/**
 * @fileoverview Framing helpers that render upstream-controlled free text as
 * clearly-delimited DATA in MCP `content[]` output. PubChem descriptions,
 * summaries, interaction statements, synonyms, and assay/GHS text are relayed
 * verbatim into the model's context; rendered as bare prose they blur the
 * boundary between server-authored instructions and retrieved data and can
 * inject markdown structure. These helpers keep that boundary explicit and
 * neutralize markdown-structure injection. `structuredContent` always carries
 * the raw value — this only shapes the human/LLM text-rendering surface.
 * @module mcp-server/tools/definitions/untrusted-text
 */
/**
 * Frame a single upstream value for safe inline interpolation into `content[]`
 * markdown. Collapses embedded newlines (so the value cannot start a new
 * markdown block) and escapes emphasis / inline-code delimiters. Proportionate
 * by design: the goal is a data boundary, not heavy escaping.
 */
export declare function inlineData(value: string): string;
/**
 * Frame multi-line upstream prose as a markdown blockquote — an explicit
 * "quoted data" boundary distinct from server-authored instructions. Every
 * line (blank ones included) is prefixed with `>` so the payload cannot break
 * out of the quote; inline emphasis/code is escaped and a leading block marker
 * (heading, quote, bullet, ordered list) is defused so it cannot restructure
 * the surrounding document.
 */
export declare function quoteData(text: string): string;
/**
 * Wrap raw upstream text in a fenced code block whose fence is guaranteed
 * longer than any backtick run inside the payload, so a stray ``` cannot close
 * the fence early and let following lines render as markdown. Preserves the
 * content verbatim (no character substitution) — used for raw V2000 SDF.
 */
export declare function fencedData(content: string, language?: string): string;
//# sourceMappingURL=untrusted-text.d.ts.map