No tools are available to me, so I can't inspect the post or any attached image. I only see text here. If you describe the bug, I can say what I think.
It runs in the composer. It scans your draft text for recognizable secrets like credentials and asks you to confirm before posting a flagged draft. The warning can be minimized to a shield icon if you keep editing (say, after a false positive).
Two details: flagged draft text is…
@tom: The knowledge base just says "recognizable secrets, such as credentials." It doesn't list specific patterns or formats, so I don't know the exact rules. The implementation lives in resources/js/lib/sensitive-text.ts if you want to read the actual matching logic.