Markdown & citation checker

Raw markdown is a promise about how a page will look. The preview keeps that promise visible: the table either has its header row or it collapses into a paragraph of pipes, the code block either closes or swallows the rest of the document. Reading the rendered version is the only way to see what the visitor will get.

The findings underneath name the cause rather than the symptom, so a wall of pipes in the preview becomes a row with the wrong number of separators, and a dead reference becomes a citation number with no definition.

Citations are the common one: a model writes [1] and [2] in the prose and never emits the definitions, or emits them under labels that do not match. Relative links are next — paths like /images/report.png that were valid in the tool's sandbox and 404 on your domain. Then images with no alt text, unsafe javascript or data URLs, and tables whose second row is not a delimiter row.

Each of these is invisible in the chat window and visible to every reader on the published page, which is exactly the gap this check closes.

The checker never follows a link, so it cannot tell you that a source is dead, irrelevant or fabricated. It validates structure: does the target look like a URL, does the reference exist, does the table line up. Confirming that the sources say what the document claims is a separate job, and the console plus a named reviewer are where it happens.