Capstone

Agent-Assisted Code Review · ~9 min

The Review Decision Table

Ten lessons, one reflex: name the review pathology in front of you, then reach for the move that resolves it — not a vaguer instruction to "review harder."

Why this, for you: review breaks down in specific, recognizable ways — a flood of agent PRs, a wall of low-value nits, a fabricated "fix," a security hole nobody traced. The table below maps each symptom you're actually staring at to the playbook move that answers it, and often the answer is "remove the PR from the queue," not "comment more." That reflex is what this course exists to make automatic.

The through-line of the whole course: review throughput, not generation, is now the bottleneck — so spend reviewer attention where signal says it matters and stay silent everywhere else. Every move below is a way to buy signal back from volume.

1 Symptom → move

Symptom you're staring atMoveFrom
PRs are piling up faster than anyone can read themTreat review throughput as the bottleneck — measure value merged, not PRs opened; cut volume before refining the checklistL1
The AI reviewer floods every PR with nitsSet a confidence floor (e.g. ≥90%) and let silence be a valid output — 29% of Copilot reviews say nothingL2
"Find bugs" prompt gives shallow, random findingsBuild a pipeline over the diff — classify, retrieve, reason, reproduce — not one LLM callL3
Every PR gets the same deep (or shallow) passTiered review — route depth by risk and tune effort; calibrate the threshold, don't review everything hardL4
The reviewer wastes attention on cosmetic hunksLabel each hunk by what it touches so attention lands where structure says it mattersL5
One big reviewer misses whole defect classesRun several specialized reviewers in parallel, then synthesize one verdict — diversity beats one promptL6
Plausible findings turn out false on inspectionGate every finding behind a reproduction step so fabricated ones never reach the humanL7
Humans and agents step on each other's workSplit by comparative advantage — agents triage breadth; humans judge intent and trade-offsL8
A 600-line agent PR with an empty descriptionAuthor for reviewability — slice small, describe the why, predict where review will choke; request a smaller PR past ~400 linesL9
The same nit gets written on every PRTurn recurring comments into learned rules so the nit is never written twiceL10

2 The one rule under all of it

Reviewing an agent-authored PR verifies context, not correctness — the code already parses and passes its own tests. The defects that ship are the ones the agent could not catch itself: gamed CI, reimplemented utilities, untested boundary cases, fabricated contracts. Every move spends scarce reviewer attention on that asymmetry, and nowhere else.

The early lessons set the economics: AI raises PR volume faster than value, so the scarce resource is the reviewer's fixed budget of attention per PR. The middle lessons spend that budget well — an inspection pipeline that runs CI changes first (the cheapest hard stop) and demands evidence last, structure labels and a committee that point attention at the right hunks, and a reproduction gate that throws out plausible-but- false findings before a human ever sees them. The late lessons protect the budget over time — author PRs to be reviewable, and fold every recurring comment back into a rule. One discipline, applied at each layer: signal over volume.

More comments is not a better review

Every backfire in this course rhymes. A wall of low-confidence comments trains you to ignore all AI output, burying the one critical finding. Each extra reviewer comment on an agent PR correlates with a 2.8-point drop in merge probability — read as required corrections, not productive alignment — so a checklist that surfaces ten findings amplifies ghosting. Past ~400 lines per diff, sustained review degrades to surface scanning regardless of checklist quality. When the checklist becomes a comfort blanket, removing PRs from the queue beats refining it.

# The reviewer's reflex, as a routing rule — not "review harder" IF diff > ~400 lines OR purpose not one sentence OR body empty: request a smaller PR # restructure is cheaper than rushing ELSE: run the order — CI first, dup scan, critical path, security, evidence last; surface only CRITICAL/HIGH findings, cluster repeats

↪ Your win: a review reflex

Mixed review — across the whole course

Question 1 · from L7Reviewing an agent-authored PR mainly verifies…

Question 2 · from the playbookIn the ten-minute inspection order, the step that comes first is…

Question 3 · from L9When a diff exceeds roughly 400 lines, the right move is to…

Question 4 · from the tableOn an agent PR, each extra reviewer comment correlates with…

Question 5 · spaced recall from Signal Over VolumeAn AI reviewer that has nothing high-confidence to add should…

You finished the course. Ask me to apply the decision table to a real review pipeline of yours, draft a tiered router with a confidence floor and a reproduction gate, or turn your last ten recurring review comments into learned rules. Or revisit any lesson — the moves compound when you use them together.
✎ Feedback