Before your invocation, the only item in the requested audit scope was the **humanizer catalog listing**. Its first line was:
> humanizer: Rewrite AI-sounding text so it reads like the writer without changing what it says.

That was a catalog description, not loaded skill content.

Your subsequent native invocation supplied the full **humanizer** skill:
- File: `<private-path>
- First line: `---`
- Public title: **Humanizer: remove AI writing patterns**
- Rule 26 title: **Re-explaining what the reader knows**

No other user instruction files, user-installed plugin or hook messages, user writing skills, memories, or Grounded Copy text were loaded in the context available to me. No source text was rewritten.