Code Review

Requesting Code Review

Run the bundled review workflow to fan out independent review dimensions over a diff, adversarially verify each finding, and consolidate the survivors — without any persistent team.

Core principle: Review early, review often. Verify findings before reporting them so false positives don't waste fix cycles.

When to Request Review

Mandatory:

• Before merge to main
• After completing a major feature

Optional but valuable:

• When stuck (fresh perspective)
• Before refactoring (baseline check)
• After fixing a complex bug

Two Modes

Within build-flow

When work is being executed by kit:build-flow, the batch review gate inside build.workflow.js already runs spec + quality review (with a staged fix-loop) at every batch boundary. Do not launch a separate review mid-build — pass the decisions ledger so reviewers have cross-batch awareness. Use a standalone review only for a final pre-merge pass.

Standalone

When you want an independent review of a diff:

1. Pick the diff ref.

• Uncommitted work (default): the workflow diffs against main.
• Committed/PR range: pass diffRef: "{BASE_SHA}..{HEAD_SHA}".

1. Locate the workflow. This skill is loaded with its base directory; the runner is /review.workflow.js.
2. Launch it:

• Workflow({ scriptPath: "/review.workflow.js", args: { diffRef, reviewDims, ledger } })
• If scriptPath rejects a bundled path, read the file and pass its contents as inline script.
• It fans out one reviewer per dimension, verifies each finding, and returns { findings }.

1. Act on the consolidated findings (below).

The default dimensions are correctness, quality, tests, security, and architecture; override reviewDims to focus or extend the review. Quality, security, and architecture run on Opus (xhigh); correctness and tests on Sonnet (high).

Acting on Findings

• Fix Critical issues immediately.
• Fix Important issues before proceeding.
• Note Minor issues for later.
• Push back if a finding is wrong — with technical reasoning and code/tests that prove it. (The workflow already verifies findings adversarially, but it is not infallible.)

Cross-Batch Awareness via the Ledger

Reviewers are ephemeral, so cross-task awareness comes from the decisions ledger the orchestrator maintains, not from a long-lived teammate's memory. Pass ledger in args so reviewers can spot inconsistencies with earlier batches ("Task 1 used X, Task 3 uses Y"), track cumulative coverage, and flag architectural drift.

Red Flags

Never:

• Skip review because "it's simple".
• Ignore Critical issues.
• Proceed with unfixed Important issues.
• Argue with valid technical feedback.

If a finding is wrong:

• Push back with technical reasoning.
• Show code/tests that prove it works.

Integration

• kit:build-flow — Its batch review gate covers in-flight batches; this skill handles standalone/pre-merge reviews.
• Review rubric: agents/code-reviewer.md — canonical review framework (plan alignment, code quality, architecture, documentation), and code-review/code-reviewer.md — context placeholders and output format that inform the workflow's dimension prompts.
• Workflow: ./review.workflow.js — the bundled fan-out-and-verify runner this skill launches.

Scope and Recommended Focus

This skill's unique value is plan-alignment review (verifying implementation matches requirements and design intent) and workflow-based review orchestration (fan-out dimensions + adversarial verification). General code quality review — style, linting, best practices, vulnerability scanning — should be deferred to the official code-review and pr-review-toolkit plugins, which provide multi-agent pipelines and specialist depth purpose-built for those concerns.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: shousper
• Source: shousper/claude-kit
• License: MIT

Install and usage instructions live in the source repository linked above.