Markdown Code Block Bundle Exporter

How to use Markdown Code Block Bundle Exporter

Recover examples from original Markdown files without copying fences, accidentally treating inline code as a program, or trusting filenames in an info string. The tool collects actual CommonMark block tokens, keeps both normalized code content and exact original full-line slices, and creates a usable bundle of inert .txt members plus an evidence manifest. No selected code or HTML is run, rendered or fetched.

1. Choose one to five actual UTF-8 Markdown files, or load the original local example file.
2. Choose all code blocks, fenced blocks or indented blocks and run the bounded extraction.
3. Review the language/info strings, normalized content and exact source-line evidence; download the ZIP and inspect its .txt members before reusing any code.

Good to know

• At most five files, 200,000 aggregate bytes, 50,000 UTF-16 source code units and 5,000 physical lines. Strict UTF-8 and well-formed Unicode are required; literal NUL is refused. One leading BOM is ignored only for parsing and remains in original source-line slices. File names are manifest text, never archive paths.
• Pinned CommonMark parsing uses Markdown-it 15.0.3 without plugins, MDX compilation or GitHub-specific extensions. Inline code spans and HTML pre/code elements are not extracted as code blocks. Fence language/info strings are evidence text and never interpreted as filenames, executable types or commands.
• Code content follows parser normalization: LF line endings and a canonical trailing LF for nonempty content. It is not a byte-for-byte extraction of original code indentation or line endings. Exact full source-line slices retain original line endings, BOM when included, list/quote prefixes and fence text. Line numbers are one-based and inclusive; original UTF-16 offsets are zero-based and end-exclusive.
• The native module Worker has a separate 10-second parser-startup cap and a 2-second computation cap, then terminates. At most 10,000 aggregate tokens, 40 levels and 200 code blocks are accepted. Excessive source is refused without a partial bundle. A current native Worker and a working local download are required.
• All modes share the same source-order extraction workflow. The ZIP contains generated block-NNN.txt files and manifest.json, even when no blocks match. The manifest retains original code-source slices and file names; it is not an anonymization or credential-removal tool. Archive text is capped at 2,000,000 UTF-8 bytes and the readable report at 1,000,000 characters.
• Processing is local. Output is limited to 1,000,000 text characters unless a stricter file limit is stated.