mbox is pre-1.0. Security fixes are applied to the latest published 0.x release only; please upgrade to the newest version before reporting.

mbox is intended to parse and render structured text; hosts that load or show untrusted documents must define the resource and navigation policies around it.

• mbox provides no input-size, CPU, or interruption-limit API. Its document cache reparses the entire document after each edit. GFM parsing goes through mdast-util-from-markdown / micromark and can be superlinear on malformed input with many unclosed delimiters. Inline parsing also temporarily allocates scan indexes proportional to an input block (about 13 B/character for a 3.6 MB single line), released after parsing. Set limits at fetch, decode, paste, and programmatic-update entry points, and use a worker when needed.
• image: { renderMode: "preview" } can load image URLs specified by the document. In a browser, <img> requests mainly create external-communication concerns such as IP, user-agent, or referrer tracking, bandwidth use, and image-decoding resource consumption. They can also become SSRF risks when a host performs server-side rendering, uses an image proxy, or fetches from an internal network. Use the default "token" mode unless external image loading is intended, and enforce appropriate URL, network, CSP, and privacy policies.
• GFM link destinations, including linkUrl return values, are revalidated against the allowlist and permanently blocked protocols. Host-supplied pageUrl, iconUrl, idUrl, Scrapbox image: { resolveSrc }, and mbox or Scrapbox completion-candidate iconUrl values reach href or src without a scheme check; enforce the host's policy. Document image URLs from GFM, mbox, and Scrapbox likewise reach img.src unchecked. By contrast, Scrapbox image click links and document-derived Scrapbox or Hatena links have parser scheme restrictions.
• For untrusted documents, prefer an explicit language allowlist to codeHighlight: { languages: "all" }. Loading many language chunks can affect network use, load volume, and predictability; this is about controlling those costs, not a claim of arbitrary code execution.

Please do not open a public issue for security problems.

Report privately through GitHub's private vulnerability reporting: go to the repository's Security tab and choose "Report a vulnerability" (https://github.com/synpuls/mbox/security/advisories/new). This keeps the report confidential until a fix is available.

Include, where possible: affected version, a description of the issue, and a minimal reproduction. We aim to acknowledge reports within a few days and will coordinate a fix and disclosure timeline with you.