Code review checklist
Make security (injection, auth), hotspots, testability, and observability explicit so review is not just style debate—tune per stack and refresh after incidents.
Category · Code review
Code review
5 skills Category 3 of 20
This category productizes human + agent review: executable checklists, automation in CI alongside bots, pairing roles, rubber-ducking assumptions, and safe refactors without behavior change. Layer with collaboration & tooling and testing to catch issues before merge.
In the hub it sits in the “code & collaboration” band; compare with quality-focused entries in the case hub. The five cards match the main hub.
Quick links
In depth
Review automation
Run static rules and agent summaries on pull requests, flag risky diffs and suggested tests—complement human review; tune noise so alerts stay actionable.
Pair programming
Agree Driver/Navigator or “human steers, agent drafts” cadence with small commits and frequent tests—good for large refactors or onboarding in unfamiliar repos.
Rubber duck debugging
Explain repro steps and hypotheses to a person or agent to surface gaps in logic—pairs with minimal repro and log redaction in observability.
Safe refactoring
Mechanical changes (rename, move, extract) under tests or contracts, each step revertible—keep behavior unchanged and diffs readable for review and bisect.