yellow-review

Parallel resolution of unresolved PR review comments. Use when you want to address all pending review feedback on a PR by spawning parallel resolver agents.

Sequential review of multiple PRs — your Graphite stack, all open PRs, or a single PR. Use when you want to review an entire stack in dependency order or batch-review all your open PRs.

Adaptive multi-agent review of a single PR. Use when you want comprehensive code review with automatic agent selection based on PR size and content.

Validate review command prerequisites and optional yellow-core integration. Use when first installing the plugin or when review commands fail before analysis begins.

General code review for correctness, CLAUDE.md compliance, and project conventions. Use when reviewing any PR for code quality, style violations, logic errors, or guideline adherence. Always selected during adaptive review.

Post-fix simplification pass (pass 2) preserving all functionality. Use when reviewing PRs after other review agents have applied fixes, to identify remaining unnecessary complexity, redundant abstractions, and YAGNI violations. Runs as the final review pass. For pre-fix complexity analysis, see code-simplicity-reviewer (yellow-core).

Code comment accuracy and rot detection. Use when reviewing PRs that add or modify documentation comments, docstrings, JSDoc, or inline comments to verify they accurately reflect the code they describe.

Test coverage and behavioral completeness analysis. Use when reviewing PRs that include test files or add testable business logic to verify tests cover critical paths, edge cases, and failure modes.

Silent failure and error handling analysis. Use when reviewing PRs that contain try-catch blocks, error handling, fallback logic, or any code that could potentially suppress errors to ensure failures are visible and actionable.

Type design, encapsulation, and invariant analysis. Use when reviewing PRs that introduce or modify type definitions (interfaces, classes, structs, enums, models) in TypeScript, Python, Rust, or Go to ensure strong invariants and proper encapsulation.

Implements fixes for individual PR review comments. Use when spawned in parallel by /review:resolve to address a single unresolved review thread by reading the file, understanding the comment, and applying the requested change.

Internal reference for PR review workflow patterns. Use when agents or commands need shared conventions for adaptive selection, output format, or error handling.