code review

"We introduced Claude Code subagents and refreshed our review system with specialized agents, and the quality of the review improved surprisingly."

"Code tasks show Claude to be the most reliable."

"Claude Code's plan mode is powerful. You enter a read-only research environment with Shift+Tab twice, analyze the codebase, and then create an implementation plan. Because you can check 'what to change and how' before implementation, rework is greatly reduced."

"The conclusion is that with a 10-minute setup, you can get a review just by writing @claude in a PR."

"The two new skills added to Claude Code, /simplify and /batch, automate these two pains."

"This article details how to leverage Claude Code with GitHub Actions."

"PRのURLを渡すだけで、以下が自動で出力されるレビュー支援ツールが欲しかった。"

"Simple apps → AI is sufficient; Complex apps → AI generation + human review is optimal."

"The reviewer stated, "The AI-generated code is more organized than code written by a new hire.""

"To prevent the 'it seems good' approval, understanding checklists are incorporated into the MR template."

"Claude CodeのPlan Mode終了をフックで捉え、PlanをCodex CLIまたはGemini CLIに送ってレビューさせるCLIツールです。"

"Code review has transformed from mentoring into "AI code inspection work.""

"In this method, AI is separated into the following five independent roles: PM/Scrum Master, Coding Agent, Code Reviewer, System Auditor, and User/Operational Support."

"AI-native development is a development method that incorporates AI as a member of the development team and structurally redesigns the roles of humans and AI."

"This article is written for those engineers who have started using AI in their daily work but are not utilizing it effectively, and those who want to optimize tasks such as code review, debugging, and design."

"By adhering to these three principles, the practical use of Claude Code saw a two-level increase in perceived accuracy."

"This article practically explains how to write code that doesn't get pointed out, with a shift-left strategy combining CLAUDE.md, Sub-Agent, SKILL.md, and verify.sh."

"In the AI era, the priority shifts: it's more effective to enhance 'validation ability (the ability to spot errors)' before focusing on implementation skills."

"The most effective combination is this: test-driven development + automated gates."

"Claude Code Action is a mechanism that allows you to run Claude on GitHub Actions workflows and automate development tasks based on issues and pull requests."

"This article introduces how Claude Code Action is used to streamline pull request operations."

"This article is the final episode (episode 6) of the "Terraform × Claude Code" series."

"In this article, we will create two sub-agents to answer this question."

"The system implements a review and feedback loop, where the coding agent's implementation is automatically checked by an AI reviewer, with a score-based correction-and-re-review loop (up to 3 times)."

"Skills版の最大の利点は、検出ルールを自分で制御できること。自社固有のセキュリティ要件に合わせたカスタマイズが可能で、ルールの追加・更新も自分のペースでできる。"

"This article discusses the 'fun things,' 'troubles,' and 'surprises' experienced while writing code with Generative AI."

"Results: Found more than 500 unknown, high-severity vulnerabilities."

"Claude can now start dev servers and preview your running app right in the desktop interface."

"I ended up building a framework to manage it and implemented it on the company i work with: comprehension checkpoints before accepting AI output, a PR template that requires explaining AI code in your own words, code review guardrails designed for AI-generated code, and a quarterly audit system."

"I wanted to create a compact template that summarizes the key points."