Text in an unread package part
A part inside the OOXML package carries text that the format's parser never opened, so its contents were not analysed.
How the text is hidden
An OOXML file is a zip of XML parts, and each format's parser opens a named set — body, headers, footnotes and comments for Word; slides, masters and notes for PowerPoint; sheets and shared strings for Excel; and, in all three, the chart, SmartArt and drawing parts under charts/, diagrams/ and drawings/. Anything outside that set is still in the package and is still read by tools that walk every entry: a part in a directory no reader claims, or a text-bearing part filed somewhere the parser does not look. Nothing has to be concealed for this to work — the text simply sits somewhere the reader does not look.
Why a model still reads it
An extractor that iterates zip entries, rather than following the format's part relationships, picks the text up verbatim. A reviewer opening the file in Word or PowerPoint sees only what the application chooses to render, which need not include that part at all.
What we do about it
For PowerPoint and Excel the scan compares every zip entry against the set of parts the parser actually opened, inflates what is left, and tests it for text markers — <w:t>, <a:t>, <c:tx>/<c:v>, plus the unqualified <t>/<is> bounded on both sides so <tc> and <tbl> do not match. Parts under media/, fonts/ and embeddings/ are excluded because each already has its own finding. Word runs its own equivalent scan inside `. The result is one informational disclosure listing the part names, action: "disclosed" with chars: 0 — nothing is stripped and nothing is flagged, and certified` goes false.
How often it fires
Not measured. We publish a rate only where one has been measured against a named corpus of the right format, and no figure exists for this technique yet. Absence means unknown — never low.
Check your own file
Three commands: a key, credit, a verdict.
Start with the API