Key points
- Self-review limitation: the generating session retains its reasoning context and is less likely to question its decisions.
- Independent review instances catch subtle issues better than self-review instructions or extended thinking.
- Multi-pass review: per-file passes for local issues, plus a cross-file integration pass for data flow between modules.
- Single-pass review of many files causes attention dilution and contradictory findings.
- A verification pass with self-reported confidence per finding enables calibrated routing to humans.
Skills the exam expects
- Review generated code with a second Claude instance that lacks the generator's reasoning.
- Split large multi-file reviews into focused per-file passes and a separate integration pass.
- Run verification passes that attach confidence to each finding for review routing.
Common misconceptions
- "Telling the same session to be its own harshest critic is as good as a second reviewer." Independence, not wording, is what helps.
- "A bigger context window fixes uneven review depth." Attention, not capacity, is the problem.
- "Per-file passes catch everything." Integration bugs live between files.
Watch for
- Generate-then-review inside one session.
- Consensus filters that drop findings caught only intermittently.
- Severity used as a stand-in for confidence when routing review.
Practice task 4.6
55 practice items cover this task in CCA Prep Studio: flashcards, matching, scenario questions and explain-it-back exercises, each tracked toward your mastery score.