CLAUDE CODE AGENTS: FUNDAMENTALS - Chapter 11 (Capstone), Exercise 2
Why Upgrading an Old Banner Format Is Refactoring, Not Coding
====================================================================================

QUESTION:
Explain why Step 6 (upgrading an old banner comment format) is described as
a refactoring task rather than a coding task, using this course's own
Chapter 8 distinction.

SOLUTION / EXPLANATION:

Chapter 8 draws a specific distinction between these two agent types based
on their success criterion: a coding agent (Chapter 5) succeeds by
producing new, correct behavior - something the code or content didn't do
before, now working correctly. A refactoring agent succeeds by the opposite
standard: the actual content and meaning must stay exactly, provably the
same, with only structure or formatting changing.

Upgrading an old single-line course-header comment to the project's current
fuller banner format fits the refactoring definition precisely. The
lesson's own actual content - its course name, chapter topic, and
educational material - doesn't change at all; only the formatting of the
metadata describing it changes, from one comment style to another. Nothing
new is being added to what a reader of the lesson actually learns, and
nothing about the lesson's own behavior or meaning is different before and
after - exactly Chapter 8's own "no observable difference, only structure
changes" standard.

If this were treated as a coding task instead, there would be a real risk of
scope creep: a coding agent's whole framing is producing new, improved
behavior, which could tempt it into also "improving" the lesson content
itself while it's already in the file changing the banner - exactly the
kind of unrequested change Chapter 5's own tip box warns against. Framing
this explicitly as a refactoring task, with the explicit constraint that
lesson content must stay identical, keeps the change scoped to exactly what
was asked: the banner format, and nothing else.

--------------------------------------------------------------------------
WHY THIS WORKS AS AN ANSWER:
It applies Chapter 8's own success-criterion distinction (new behavior vs.
identical behavior) directly to this specific task, and explains the
concrete risk (unrequested content changes) that framing it as a coding
task instead would introduce.