Challenge 2 — Solution

Task: A user sends the prompt "flesh out imagemagick1" for a topic
that's never been mentioned before. Based on P9's own description,
what should Claude actually produce in response — and what should it
NOT produce yet?

What Claude SHOULD produce: a full course outline for ImageMagick -
deciding how many courses the topic genuinely warrants (1, 2, or 3,
based on real scope, not a default), a full chapter list for each of
those courses matching the depth already used elsewhere in the bucket
list, and the folder/prompt-prefix convention the course would
eventually use once real chapters are generated - written directly
into course_bucket_list.md, replacing whatever one-line entry (or lack
of one) currently exists for this topic.

What Claude should NOT produce yet: any actual chapter CONTENT at all.
P9 is explicitly lighter-weight than the normal chapter-1 workflow -
the outline itself IS the deliverable being reviewed, and P9 doesn't
require a separate confirmation round-trip before writing it, since
producing the outline is already the reviewable output. Generating
"imagemagick1-1" as a real chapter would be jumping ahead of what was
actually asked for - that only happens later, once someone explicitly
requests that specific chapter by its own prompt (e.g.
"imagemagick1-1"), the same way rules1-1 through rules1-3 were each
requested individually after this course's own outline already existed.

Notes:
- This is exactly the same sequence that produced the very course this
  chapter belongs to: "flesh out rules1" first produced only the
  10-chapter outline in course_bucket_list.md, and it was only after
  that step that "rules1-1", "rules1-2", and "rules1-3" were each
  separately requested to generate real chapter content.
- The one exception worth noting: if the topic doesn't obviously belong
  to an existing Subject, P9 doesn't say to guess silently - clarifying
  which Subject a genuinely new, ambiguous topic belongs to is still
  reasonable before writing the outline.