/ce-code-review

/ce-code-review — “Structured code review with skill-local personas, confidence-gated findings, four modes.” Step 5 of the loop and the wiki’s richest review fan-out: ~16 skill-local reviewer personas (correctness, security, performance, maintainability, reliability, testing, api-contract, data-migration, adversarial, agent-native, deployment-verification, project-standards, previous-comments, learnings-researcher, + Swift/iOS and frontend-races stack reviewers) run in fresh context (pattern-fresh-context-subagents, pattern-parallel-persona-review), with an adversarial and optional cross-model pass (pattern-adversarial-review). Findings are confidence-gated — only those above a bar surface — and one persona (learnings-researcher) pulls the docs/solutions/ compound corpus in, so past lessons sharpen this review (pattern-knowledge-compounding on the consuming side). → artifact-review-report.

It implements stage-review (quality gate) and is the code-side sibling of ce-doc-review.

Cross-framework equivalents

Code-review cluster: ce-code-review ↔ addy-code-review ↔ bmad-code-review — multi-lens pre-merge quality review. (GSD’s gsd-verifier is validate-side too but goal-backward/spec-conformance, so it sits in the openspec-verify cluster, not here.) ce-code-review most resembles bmad-code-review (adversarial parallel layers) and Addy’s /ship persona fan-out, but adds confidence gating, cross-model review, and compound-corpus grounding.

See Also

• addy-code-review · bmad-code-review · gsd-verifier — code-review-cluster counterparts.
• addy-code-reviewer — Addy’s Staff-Engineer review persona (the single-persona analogue).
• ce-doc-review — the plan-side sibling; ce-simplify-code — runs just before this.
• stage-review — the canonical stage this implements.