CLAUDE CODE AGENTS: FUNDAMENTALS - Chapter 1, Exercise 3
Why "The Bug We Were Just Discussing" Fails as a Brief
====================================================================================

QUESTION:
A colleague launches a subagent with the brief "fix the bug we were just
discussing," expecting it to already understand which bug that refers to.
Explain why this will likely fail, using this chapter's own warning box.

SOLUTION / EXPLANATION:

This chapter's warning box is direct on exactly this point: a subagent
starts cold, with no automatic access to whatever conversation happened
before it was launched. "The bug we were just discussing" only means
something to someone who was actually present for that earlier
conversation - it refers to shared context that exists in the colleague's
own conversation with the default assistant, not in the subagent's own,
separately-started context.

When the subagent receives this brief, it has no bug description, no file
location, no symptom, and no prior discussion to draw on - just a reference
to something it has no way of resolving. This isn't a shortcoming of the
particular subagent involved; it's the direct, predictable consequence of
treating a subagent as though it shares memory with the default assistant's
conversation, which this chapter is explicit that it does not.

The likely outcome is one of two failure modes: the subagent either asks
for clarification (if that's possible in how it was launched), or it makes
an assumption about which bug is meant and works on the wrong thing
entirely - producing a result that looks like effort was made, but doesn't
address whatever the colleague actually needed fixed.

The fix, per this chapter's own point (which Chapter 3 covers in full), is
writing a self-contained brief: describing the actual bug directly -
symptoms, relevant file(s), what's expected versus what's happening -
rather than referring back to context the subagent was never given access
to in the first place.

--------------------------------------------------------------------------
WHY THIS WORKS AS AN ANSWER:
It connects the specific failure directly to the chapter's own stated fact
(a subagent starts without access to prior conversation), explains
concretely why "the bug we were just discussing" is meaningless to an agent
with no such access, and names the actual fix (a self-contained brief)
rather than treating the failure as an unexplained fluke.