Artifact format reference

Offprint HTML is the canonical capture artifact. PDF, Markdown, ZIP, self-extracting HTML, and Multipurpose Internet Mail Extensions HTML (MHTML) are export formats derived from a verified HTML source.

Choose a format

Format	Output shape	Open with	Executable content	Provenance
Offprint HTML	One .html file	Modern browser	Exact Offprint restoration programs	Full embedded artifact manifest
PDF	One .pdf file	PDF reader	Rejected by verifier	Selected document metadata, source provenance, and resource summary
Markdown	Directory with index.md and assets	Markdown reader	Raw HTML rejected	Source, capture time, and policy digest front matter by default
ZIP	One .zip file	Archive tool	Contained Offprint HTML owns its normal restoration programs	Full embedded and sidecar manifests
Self-extracting HTML	One .compressed.html file	Browser with DecompressionStream	Inline Offprint decompression loader	Full manifest inside compressed Offprint HTML
MHTML	One .mhtml file	Browser with MHTML import	Restoration execution removed	Full embedded and sidecar manifests

Offprint HTML

The HTML artifact embeds render dependencies, the artifact manifest, exact state-restoration programs, and a matching content security policy. Use artifact verify for static or offline verification. HTML cannot be requested through FormatSpec or verify_format.

PDF

PDF export prints the offline-verified page through Chromium. It preserves selectable text, printable HTTP, HTTPS, mail, and telephone links, tagged structure, document outline, and selected source metadata.

Offprint activates print styles before offline verification and loads and decodes HTML image elements, including lazy images in shadow trees and inline frames, before printing. Linked stylesheets retain their media and disabled state.

Export first verifies the source HTML in screen media, then verifies the PDF's print layout with deferred images loaded. Rust's export_capture reuses the capture receipt's screen verification. Links to captured element IDs become internal PDF destinations. Other links resolve against the captured source URL. Follow the PDF guide for a complete capture and export workflow.

The verifier rejects encryption, embedded files, rich media, local file links, automatic actions, executable actions, malformed content streams, and metadata that disagrees with the Offprint provenance packet.

The provenance packet preserves capture warning codes, including repeated occurrences, within 16,384 entries and 1 MiB of warning text. Exceeding either limit fails before the PDF is committed.

Options:

• landscape prints in landscape orientation.
• preferCssPageSize uses the captured document's CSS page size when present.

Markdown

Markdown export preserves a semantic subset of the document, including headings, paragraphs, emphasis, code, links, images, blockquotes, lists, tables, and inline frame content.

Images become content-addressed files under assets/. The verifier rejects raw HTML, unsafe link schemes, missing assets, extra assets, and invalid asset digests. Front matter contains source, capture time, and policy digest unless frontMatter is false. The Markdown verifier validates syntax, links, and asset integrity. It does not validate front-matter provenance values.

Markdown encoding accepts at most 10,000 unique assets. Its text buffers and asset content share the remaining byte budget for the export request.

ZIP

ZIP contains exactly two ordered deflated members:

index.html
offprint-manifest.json

The verifier checks the archive structure, statically verifies index.html, and requires the sidecar manifest to match the embedded manifest exactly.

Self-extracting HTML

Self-extracting HTML stores gzip-compressed Offprint HTML in a base64 payload. An inline loader uses the browser's DecompressionStream API and document.write to replace the shell with the decompressed artifact.

This export executes the known loader and uses script-src 'unsafe-inline' in its shell. Treat it differently from canonical safe-static HTML. The verifier checks the exact shell bytes, compressed payload digest, decompression limit, and static validity of the contained Offprint HTML.

MHTML

MHTML is a deterministic multipart/related message with two base64 parts:

• Sandboxed HTML
• offprint-manifest.json

The encoder removes Offprint restoration execution before packaging. An artifact that requires structural repair cannot be represented as MHTML. The verifier checks the exact two-part structure, manifest equality, content locations, deterministic boundary, and static sandboxed HTML.

Fixed verification limits

• HTML inspection and source loading: 64 MiB
• Decoded HTML inside container formats: 64 MiB
• Decoded manifest sidecar: 16 MiB
• All formats in one export request: 256 MiB and 10,001 files combined
• Markdown assets: 10,000 unique files, within the request's file budget

Each completed format is charged before Offprint retains it for delivery. Markdown also checks its byte budget while building text and decoding assets. If a limit is exceeded, the export fails before committing output files.

Format verifiers establish the representation-specific structure and content contract. ZIP and MHTML verify full manifest sidecars, PDF verifies its selected provenance projection, and self-extracting HTML verifies the contained Offprint HTML. Markdown provenance remains emitted metadata. Format verification does not perform a network-denied browser reopen.