What problem does it solve?
Feature work in the Civitai main Next.js app (src/) often ships without a structured review covering reuse, correctness, performance, tests, and intent, letting defects like N+1 queries, auth scoping gaps, and vacuous tests reach production.
Core Features & Use Cases
- Five parallel review lanes: Launches reuse, correctness/safety, performance, test, and intent review agents simultaneously against a scoped diff, each with Civitai-specific knowledge of services, components, and traps.
- Scoped diffing against the real base: Determines the actual integration branch base (not main) via the PR, so unrelated commits do not pollute the review.
- Consolidated three-section report: Separates intent findings, ranked code findings, and pre-existing duplication mapping, with deduplication and disproval of false positives.
- Explicit fix loop: Reviewers report and never patch; the implementer fixes, and only touched lanes re-run against the updated diff until no unresolved findings remain.
- Use Case: Before calling a feature segment done, run this skill to fan out all five agents over the segment's diff, chase any silent lanes, consolidate findings, and drive the fix loop to ship.
Quick Start
Run the civitai-review skill over the current feature segment's diff against its PR base branch and consolidate the five agents' findings before committing.
