CLAUDE CODE AGENTS: FUNDAMENTALS - Chapter 6, Exercise 1
Why a Review Agent Is Denied Edit Access Even Though It Could Fix Things
====================================================================================

QUESTION:
Explain why a code-review agent is deliberately denied Edit access, even
though it would be technically capable of fixing many of the issues it
finds.

SOLUTION / EXPLANATION:

This isn't about capability - a capable model could plausibly fix many of
the issues it notices while reviewing code. The restriction exists for a
different reason: keeping the act of finding a problem separate from the
act of fixing it, so the review itself stays honest and complete.

If a review agent had Edit access, it could quietly fix something it
noticed along the way, rather than reporting it as a finding - meaning the
person relying on the review would never actually see that the issue
existed in the first place, or get a chance to understand it, disagree with
the fix, or decide it wasn't actually a real problem. The review's whole
value is a complete, honest account of what's wrong; silently patching
issues instead of reporting them undermines that account, even if the
patches themselves happen to be correct.

There's also a structural benefit tied to Chapter 1's own isolation
principle: an agent with no Edit access is guaranteed, by construction, to
never accidentally alter the code under review while examining it - no risk
of a review agent introducing a new bug while poking around, or changing
something the developer didn't ask it to touch. The restriction isn't
distrust of the agent's competence; it's a deliberate structural guarantee
that a review stays a review, and any actual fix happens as its own
separate, visible, reviewable step - typically handed to a coding agent
(Chapter 5) afterward.

--------------------------------------------------------------------------
WHY THIS WORKS AS AN ANSWER:
It explains that the restriction isn't about the agent's competence to fix
issues, but about preserving the review's own completeness and honesty as a
separate, visible step, and ties this back to Chapter 1's isolation
principle as a structural guarantee rather than a matter of trust.