code-review

Automate parallel specification, quality, and security code reviews with prioritized findings.

Updated Mar 29, 2026
One-click install
npx skills add https://github.com/sizukutamago/claude-code-harness --skill code-review-sizukutamago
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: code-review
Source: https://github.com/sizukutamago/claude-code-harness/tree/main/.claude/skills/code-review
Command: npx skills add https://github.com/sizukutamago/claude-code-harness --skill code-review-sizukutamago

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill removes the blind spots and inconsistency of ad-hoc reviews by enforcing a structured, three-perspective code review (specification compliance, code quality, security) so that implementation drift, maintainability regressions, and vulnerabilities are detected and remediated before merge.

Core Features & Use Cases

  • Parallel 3‑perspective review: Dispatches independent spec-compliance, quality, and security reviewers to analyze the same code in parallel and produce prioritized MUST/SHOULD/CONSIDER findings.
  • Dependency-aware fix workflow: Performs affected-file dependency analysis, sequences shared-layer fixes serially and independent fixes in parallel, and enforces a bounded review→fix loop with escalation rules.
  • Review memory integration: Clusters findings via review-memory tooling, appends findings to review-findings.jsonl, promotes recurring clusters into review-conventions.md, and optionally triggers product- or harness-user observational reviews.
  • Use case: After a feature implementation with all tests green and a requirements/REQ- path, run this Skill to get a consolidated, prioritized review report and a coordinated plan for implementers to apply fixes.

Quick Start

Run the code-review skill on a completed implementation with tests green and the corresponding requirements path to produce a grouped MUST/SHOULD/CONSIDER report, clustered findings, and a repair plan.

Frequently Asked Questions about code-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate code review for spec compliance, quality, and security?

Automate code review by dispatching parallel reviewers to analyze specification compliance, code quality, and security, producing prioritized MUST/SHOULD/CONSIDER findings. This structured approach detects implementation drift and vulnerabilities before merge.

What is the best way to manage code review findings and recurring issues?

Manage code review findings by clustering them with review-memory tooling and appending records to review-findings.jsonl. Recurring clusters are promoted into review-conventions.md to establish lasting conventions and coordinate implementer fix phases.

How do you sequence code fixes after identifying dependency conflicts?

Sequence code fixes by performing affected-file dependency analysis to execute shared-layer fixes serially and independent fixes in parallel. This dependency-aware workflow enforces a bounded review loop with escalation rules.

Do I need passing tests and a requirements file to run a code review?

Yes, you need all tests green and a requirements.md file available. This code review process is applicable to completed feature implementations, bug fixes, and refactors where specification paths are clearly defined.

Can I run parallel code reviews on a completed feature implementation?

Yes, you can run parallel code reviews on completed implementations. Independent spec-compliance, quality, and security reviewers analyze the same code simultaneously to produce a consolidated, prioritized report and a coordinated repair plan.

Why does ad-hoc code review miss implementation drift and vulnerabilities?

Ad-hoc code review misses implementation drift due to a lack of structured perspectives. Enforcing a three-perspective review covering specification compliance, code quality, and security removes blind spots to detect maintainability regressions and vulnerabilities.