CLAUDE CODE AGENTS: ADVANCED ORCHESTRATION - Chapter 1, Exercise 2
Why Chaining Agents Makes "Starts Cold" Worse, Not Just Repeated
====================================================================================

QUESTION:
Explain why Chapter 1's "an agent starts cold" problem is described in this
chapter as getting worse, not just repeated, once several agents are
chained together.

SOLUTION / EXPLANATION:

Fundamentals' Chapter 1 established that a single subagent starts with no
memory of the conversation that launched it - it needs a self-contained
brief to work from. In a single-agent scenario, this is a one-time concern:
you write one good brief, the agent works from it, and that's the entire
interaction.

In a chain of several agents, this same problem doesn't just happen once
more per agent - it compounds, because each agent's output has to become
the NEXT agent's input. If the first agent in a chain doesn't clearly
capture everything the next agent will need, that gap doesn't stay
contained to the first agent's own step - it propagates forward, since the
second agent starts just as cold as the first did, except now its only
source of context is whatever the first agent's own output happened to
include. A gap or ambiguity introduced at any point in the chain doesn't
just affect that one step; it can silently degrade every step that comes
after it, since each later agent has no way to independently recover
context that never made it into what was handed to it.

This is precisely why this chapter frames orchestration as needing
deliberate attention to carrying context FORWARD through a chain, not just
briefing each agent individually the way Fundamentals taught for one agent
in isolation. The problem isn't simply "the same issue, N times" - it's a
compounding risk where an early mistake can echo through every step that
follows it.

--------------------------------------------------------------------------
WHY THIS WORKS AS AN ANSWER:
It explains specifically how a context gap at one point in a chain
propagates to every subsequent agent, rather than staying isolated to that
one step, showing why the problem compounds rather than merely repeating
across a chain of otherwise-independent single-agent interactions.