{"id":76154,"date":"2026-05-20T06:21:58","date_gmt":"2026-05-20T06:21:58","guid":{"rendered":"https:\/\/www.devopsschool.com\/blog\/?p=76154"},"modified":"2026-05-20T06:22:00","modified_gmt":"2026-05-20T06:22:00","slug":"how-to-review-github-pull-requests-from-the-terminal-using-gh-and-claude-code","status":"publish","type":"post","link":"https:\/\/www.devopsschool.com\/blog\/how-to-review-github-pull-requests-from-the-terminal-using-gh-and-claude-code\/","title":{"rendered":"How to Review GitHub Pull Requests from the Terminal Using gh and Claude Code"},"content":{"rendered":"\n

<\/p>\n\n\n\n

    \n
  1. AI-Powered Pull Request Reviews with GitHub CLI and Claude Code<\/strong><\/li>\n\n\n\n
  2. Build a Simple PR Review Bot Using gh<\/code>, Git, and Claude<\/strong><\/li>\n\n\n\n
  3. Speed Up Code Reviews with Claude Code and GitHub CLI<\/strong><\/li>\n\n\n\n
  4. Review GitHub PRs Locally with Claude: A Practical Bash Script<\/strong><\/li>\n\n\n\n
  5. From PR to AI Review: Automating Developer Feedback with gh<\/code> and Claude<\/strong><\/li>\n\n\n\n
  6. Using Claude Code as Your Terminal-Based PR Reviewer<\/strong><\/li>\n\n\n\n
  7. A Practical Guide to AI-Assisted GitHub PR Reviews<\/strong><\/li>\n\n\n\n
  8. Automated PR Review Workflow with Bash, GitHub CLI, and Claude<\/strong><\/li>\n\n\n\n
  9. How I Use Claude to Review Pull Requests from the Command Line<\/strong><\/li>\n\n\n\n
  10. Claude + GitHub CLI: A Fast Workflow for Pull Request Reviews<\/strong><\/li>\n\n\n\n
  11. <\/li>\n<\/ol>\n\n\n\n

    How to Review GitHub Pull Requests from the Terminal Using gh<\/code> and Claude Code<\/h1>\n\n\n\n
    \n\n\n\n

    Introduction<\/h2>\n\n\n\n

    Code review is one of the most important parts of software development, but it can also be time-consuming. Every pull request needs to be checked for correctness, security issues, missing tests, edge cases, and maintainability concerns.<\/p>\n\n\n\n

    With tools like GitHub CLI<\/strong>, Git<\/strong>, and Claude Code<\/strong>, we can create a simple terminal workflow that helps review pull requests faster.<\/p>\n\n\n\n

    In this blog, we will build a Bash script that:<\/p>\n\n\n\n

      \n
    • accepts a GitHub PR number<\/li>\n\n\n\n
    • checks out the pull request locally<\/li>\n\n\n\n
    • collects PR metadata and diff<\/li>\n\n\n\n
    • asks Claude to review the PR<\/li>\n\n\n\n
    • writes the review to a Markdown file<\/li>\n\n\n\n
    • optionally posts the review back to the GitHub PR<\/li>\n<\/ul>\n\n\n\n

      This is not meant to replace human review. Instead, it acts like an extra reviewer that can quickly spot bugs, risks, missing tests, and suspicious changes.<\/p>\n\n\n\n

      \n\n\n\n

      Prerequisites<\/h2>\n\n\n\n

      Before using the script, your system should have the following tools installed and configured:<\/p>\n\n\n

      git --version\ngh --version\nclaude --version\n<\/code><\/span><\/pre>\n\n\n
      You also need to authenticate GitHub CLI:<\/p>\n\n\n
      gh auth login\n<\/code><\/span><\/pre>\n\n\n
      It is also useful to configure GitHub CLI with Git credentials:<\/p>\n\n\n
      gh auth setup-git\n<\/code><\/span><\/pre>\n\n\n
      This helps avoid authentication prompts when the script fetches PR branches.<\/p>\n\n\n\n
      \n\n\n\n
      Why use gh<\/code>?<\/h2>\n\n\n\n
      gh<\/code> is the official GitHub command-line tool. It allows us to work with GitHub directly from the terminal.<\/p>\n\n\n\n
      For pull request reviews, gh<\/code> gives us useful commands like:<\/p>\n\n\n
      gh pr view 56\ngh pr checkout 56\ngh pr diff 56\ngh pr comment 56\n<\/code><\/span><\/pre>\n\n\n
      Instead of manually opening GitHub in the browser, copying the diff, and pasting it into Claude, we can automate the whole flow.<\/p>\n\n\n\n
      \n\n\n\n
      Basic workflow<\/h2>\n\n\n\n
      The workflow looks like this:<\/p>\n\n\n
      PR number\n   \u2193\ngh pr checkout\n   \u2193\ncollect PR metadata\n   \u2193\ncollect changed files\n   \u2193\ncollect patch\/diff\n   \u2193\nsend context to Claude\n   \u2193\nsave review as<\/span> Markdown\n   \u2193\noptionally post review to GitHub\n<\/code><\/span>Code language:<\/span> JavaScript<\/span> (<\/span>javascript<\/span>)<\/span><\/small><\/pre>\n\n\n
      This is simple, transparent, and easy to customize.<\/p>\n\n\n\n
      \n\n\n\n
      The review script<\/h2>\n\n\n\n
      Save the following as:<\/p>\n\n\n
      review<\/span>.sh<\/span>\n<\/code><\/span>Code language:<\/span> CSS<\/span> (<\/span>css<\/span>)<\/span><\/small><\/pre>\n\n
      #!\/usr\/bin\/env bash<\/span>\nset -euo pipefail\n\nPR=\"${1:?Usage: .\/review.sh <pr-number> [--post]}\"<\/span>\nPOST=\"${2:-}\"<\/span>\n\nOUT=\"claude-pr-${PR}-review.md\"<\/span>\nCTX=\"\/tmp\/pr-${PR}-context.md\"<\/span>\nBRANCH=\"pr-${PR}\"<\/span>\nREMOTE=\"${GH_REMOTE:-origin}\"<\/span>\n\nrequire_cmd() {\n  command -v \"$1\"<\/span> >\/dev\/null<\/span> 2<\/span>>&1<\/span> || {\n    echo<\/span> \"Error: required command not found: $1\"<\/span>\n    exit<\/span> 1<\/span>\n  }\n}\n\nrequire_cmd git\nrequire_cmd gh\nrequire_cmd claude\n\necho<\/span> \"Checking GitHub authentication...\"<\/span>\ngh auth status >\/dev\/null<\/span>\n\necho<\/span> \"Checking Git remote...\"<\/span>\nif<\/span> ! git remote get-url \"$REMOTE\"<\/span> >\/dev\/null<\/span> 2<\/span>>&1<\/span>; then\n  echo<\/span> \"Error: Git remote '$REMOTE' not found.\"<\/span>\n  echo<\/span>\n  echo<\/span> \"Available remotes:\"<\/span>\n  git remote -v\n  echo<\/span>\n  echo<\/span> \"You can run with a different remote like this:\"<\/span>\n  echo<\/span> \"  GH_REMOTE=upstream .\/review.sh $PR\"<\/span>\n  exit<\/span> 1<\/span>\nfi\n\nREMOTE_URL=\"$(git remote get-url \"<\/span>$REMOTE\")\"<\/span>\n\necho<\/span> \"Reading PR metadata...\"<\/span>\nPR_STATE=\"$(gh pr view \"<\/span>$PR\" --json state --jq '.state')\"<\/span>\nBASE_REF=\"$(gh pr view \"<\/span>$PR\" --json baseRefName --jq '.baseRefName')\"<\/span>\nHEAD_REF=\"$(gh pr view \"<\/span>$PR\" --json headRefName --jq '.headRefName')\"<\/span>\nREPO_FULL=\"$(gh repo view --json nameWithOwner --jq '.nameWithOwner')\"<\/span>\n\necho<\/span> \"Repository: $REPO_FULL\"<\/span>\necho<\/span> \"Remote: $REMOTE\"<\/span>\necho<\/span> \"Remote URL: $REMOTE_URL\"<\/span>\necho<\/span> \"PR: #$PR\"<\/span>\necho<\/span> \"State: $PR_STATE\"<\/span>\necho<\/span> \"Base: $BASE_REF\"<\/span>\necho<\/span> \"Head: $HEAD_REF\"<\/span>\necho<\/span>\n\nCHECKOUT_OK=\"false\"<\/span>\n\necho<\/span> \"Checking out PR #$PR...\"<\/span>\n\nif<\/span> gh pr checkout \"$PR\"<\/span> --force; then\n  echo<\/span> \"Checked out PR using gh pr checkout.\"<\/span>\n  CHECKOUT_OK=\"true\"<\/span>\nelse<\/span>\n  echo<\/span> \"gh pr checkout failed.\"<\/span>\n  echo<\/span> \"Trying GitHub pull-request ref via local remote '$REMOTE'...\"<\/span>\n\n  gh auth setup-git >\/dev\/null<\/span> 2<\/span>>&1<\/span> || true<\/span>\n\n  if<\/span> git fetch \"$REMOTE\"<\/span> \"+refs\/pull\/${PR}\/head:refs\/remotes\/${REMOTE}\/${BRANCH}\"<\/span>; then\n    git switch<\/span> -C \"$BRANCH\"<\/span> \"refs\/remotes\/${REMOTE}\/${BRANCH}\"<\/span>\n    echo<\/span> \"Checked out PR using refs\/pull\/${PR}\/head.\"<\/span>\n    CHECKOUT_OK=\"true\"<\/span>\n  else<\/span>\n    echo<\/span>\n    echo<\/span> \"Could not checkout PR branch.\"<\/span>\n    echo<\/span> \"This can happen when:\"<\/span>\n    echo<\/span> \"- the PR is merged or closed\"<\/span>\n    echo<\/span> \"- the source branch was deleted\"<\/span>\n    echo<\/span> \"- your Git remote auth is not configured\"<\/span>\n    echo<\/span> \"- origin points somewhere unexpected\"<\/span>\n    echo<\/span>\n    echo<\/span> \"Continuing with diff-only review using gh pr diff.\"<\/span>\n    CHECKOUT_OK=\"false\"<\/span>\n  fi\nfi\n\necho<\/span>\necho<\/span> \"Collecting PR context...\"<\/span>\n\n{\n  echo<\/span> \"# PR #$PR Review Context\"<\/span>\n  echo<\/span>\n  echo<\/span> \"Repository: $REPO_FULL\"<\/span>\n  echo<\/span> \"Remote: $REMOTE\"<\/span>\n  echo<\/span> \"PR state: $PR_STATE\"<\/span>\n  echo<\/span> \"Base branch: $BASE_REF\"<\/span>\n  echo<\/span> \"Head branch: $HEAD_REF\"<\/span>\n  echo<\/span> \"Checkout available: $CHECKOUT_OK\"<\/span>\n  echo<\/span>\n\n  echo<\/span> \"## PR metadata\"<\/span>\n  gh pr view \"$PR\"<\/span> --json title,body,author,state,baseRefName,headRefName,headRepositoryOwner,headRepository,isCrossRepository,files,commits\n  echo<\/span>\n\n  echo<\/span> \"## Changed files\"<\/span>\n  gh pr diff \"$PR\"<\/span> --name-only\n  echo<\/span>\n\n  echo<\/span> \"## Patch\"<\/span>\n  gh pr diff \"$PR\"<\/span> --patch\n} > \"$CTX\"<\/span>\n\necho<\/span> \"PR context written to $CTX\"<\/span>\necho<\/span>\necho<\/span> \"Asking Claude to review...\"<\/span>\n\ncat \"$CTX\"<\/span> | claude -p \"\nYou are a senior software engineer reviewing GitHub PR #$PR.\n\nRepository:\n$REPO_FULL\n\nPR state:\n$PR_STATE\n\nCheckout available:\n$CHECKOUT_OK\n\nReview goals:\n- Find correctness bugs.\n- Find security or data-leak risks.\n- Find missing tests for changed behavior.\n- Find API compatibility issues.\n- Find race conditions, error handling gaps, and edge cases.\n- Avoid nitpicks unless they materially affect maintainability.\n\nInstructions:\n- Do not edit files.\n- Use the PR metadata and patch.\n- If checkout is available, inspect changed files in the repository if needed.\n- If checkout is not available, review from the supplied diff only.\n- If there are no serious issues, say so clearly.\n- Keep the review practical and actionable.\n- Include file paths and reasoning.\n- Clearly separate must-fix issues from suggestions.\n- End with a concise comment suitable for posting on the GitHub PR.\n\nOutput format:\n\n## Claude PR Review\n\n### Summary\n\n### Must fix\n\n### Should consider\n\n### Tests to run or add\n\n### Suggested PR comment\n\"<\/span> > \"$OUT\"<\/span>\n\necho<\/span>\necho<\/span> \"Review written to $OUT\"<\/span>\n\nif<\/span> [[ \"$POST\"<\/span> == \"--post\"<\/span> ]]; then\n  echo<\/span> \"Posting review comment to PR #$PR...\"<\/span>\n  gh pr comment \"$PR\"<\/span> --body-file \"$OUT\"<\/span>\n  echo<\/span> \"Posted.\"<\/span>\nelse<\/span>\n  echo<\/span> \"Not posted. Review the file first:\"<\/span>\n  echo<\/span> \"  less $OUT\"<\/span>\n  echo<\/span>\n  echo<\/span> \"To post it:\"<\/span>\n  echo<\/span> \"  gh pr comment $PR --body-file $OUT\"<\/span>\n  echo<\/span>\n  echo<\/span> \"Or run:\"<\/span>\n  echo<\/span> \"  .\/review.sh $PR --post\"<\/span>\nfi\n<\/code><\/span>Code language:<\/span> PHP<\/span> (<\/span>php<\/span>)<\/span><\/small><\/pre>\n\n\n
      \n\n\n\n
      Make the script executable<\/h2>\n\n\n
      chmod<\/span> +x<\/span> review<\/span>.sh<\/span>\n<\/code><\/span>Code language:<\/span> CSS<\/span> (<\/span>css<\/span>)<\/span><\/small><\/pre>\n\n\n
      \n\n\n\n
      Usage<\/h2>\n\n\n\n
      To review a PR:<\/p>\n\n\n
      .\/review.sh 56\n<\/code><\/span><\/pre>\n\n\n
      This creates a file like:<\/p>\n\n\n
      claude-pr-56-review<\/span>.md<\/span>\n<\/code><\/span>Code language:<\/span> CSS<\/span> (<\/span>css<\/span>)<\/span><\/small><\/pre>\n\n\n
      You can inspect the result:<\/p>\n\n\n
      less<\/span> claude-pr-56-review<\/span>.md<\/span>\n<\/code><\/span>Code language:<\/span> CSS<\/span> (<\/span>css<\/span>)<\/span><\/small><\/pre>\n\n\n
      After reviewing the generated comment, post it manually:<\/p>\n\n\n
      gh<\/span> pr<\/span> comment<\/span> 56 --body-file<\/span> claude-pr-56-review<\/span>.md<\/span>\n<\/code><\/span>Code language:<\/span> CSS<\/span> (<\/span>css<\/span>)<\/span><\/small><\/pre>\n\n\n
      Or post automatically:<\/p>\n\n\n
      .\/review.sh 56 --post\n<\/code><\/span><\/pre>\n\n\n
      For real team usage, I recommend reviewing the generated Markdown before posting. AI review is helpful, but it should not blindly comment on teammates\u2019 PRs without a human pass.<\/p>\n\n\n\n
      \n\n\n\n
      How the script works<\/h1>\n\n\n\n
      1. Strict Bash mode<\/h2>\n\n\n
      set<\/span> -euo pipefail\n<\/code><\/span>Code language:<\/span> JavaScript<\/span> (<\/span>javascript<\/span>)<\/span><\/small><\/pre>\n\n\n
      This makes the script safer:<\/p>\n\n\n\n
        \n
      • -e<\/code> exits when a command fails<\/li>\n\n\n\n
      • -u<\/code> fails on undefined variables<\/li>\n\n\n\n
      • pipefail<\/code> catches failures inside pipelines<\/li>\n<\/ul>\n\n\n\n
        This is a good default for automation scripts.<\/p>\n\n\n\n
        \n\n\n\n
        2. Read the PR number<\/h2>\n\n\n
        PR=\"${1:?Usage: .\/review.sh <pr-number> [--post]}\"<\/span>\nPOST=\"${2:-}\"<\/span>\n<\/code><\/span>Code language:<\/span> JavaScript<\/span> (<\/span>javascript<\/span>)<\/span><\/small><\/pre>\n\n\n
        The first argument is required:<\/p>\n\n\n
        .\/review.sh 56\n<\/code><\/span><\/pre>\n\n\n
        The second argument is optional:<\/p>\n\n\n
        .\/review.sh 56 --post\n<\/code><\/span><\/pre>\n\n\n
        If the PR number is missing, the script shows usage instructions.<\/p>\n\n\n\n
        \n\n\n\n
        3. Check required tools<\/h2>\n\n\n
        require_cmd git\nrequire_cmd gh\nrequire_cmd claude\n<\/code><\/span><\/pre>\n\n\n
        The script confirms that git<\/code>, gh<\/code>, and claude<\/code> are available before doing anything else.<\/p>\n\n\n\n
        This avoids confusing errors later.<\/p>\n\n\n\n
        \n\n\n\n
        4. Check GitHub authentication<\/h2>\n\n\n
        gh auth status >\/dev\/<\/span>null<\/span>\n<\/code><\/span>Code language:<\/span> JavaScript<\/span> (<\/span>javascript<\/span>)<\/span><\/small><\/pre>\n\n\n
        This confirms that GitHub CLI is authenticated.<\/p>\n\n\n\n
        If authentication is missing, run:<\/p>\n\n\n
        gh auth login\n<\/code><\/span><\/pre>\n\n\n
        \n\n\n\n
        5. Read PR metadata<\/h2>\n\n\n
        gh pr view \"$PR\"<\/span> --json state --jq '.state'<\/span>\ngh pr view \"$PR\"<\/span> --json baseRefName --jq '.baseRefName'<\/span>\ngh pr view \"$PR\"<\/span> --json headRefName --jq '.headRefName'<\/span>\n<\/code><\/span>Code language:<\/span> JavaScript<\/span> (<\/span>javascript<\/span>)<\/span><\/small><\/pre>\n\n\n
        The script reads useful PR details:<\/p>\n\n\n\n
          \n
        • PR state<\/li>\n\n\n\n
        • base branch<\/li>\n\n\n\n
        • head branch<\/li>\n\n\n\n
        • repository name<\/li>\n<\/ul>\n\n\n\n
          This metadata helps Claude understand the context before reviewing the diff.<\/p>\n\n\n\n
          \n\n\n\n
          6. Try normal PR checkout<\/h2>\n\n\n
          gh pr checkout \"$PR\"<\/span> --force\n<\/code><\/span>Code language:<\/span> JavaScript<\/span> (<\/span>javascript<\/span>)<\/span><\/small><\/pre>\n\n\n
          This is the cleanest way to check out a PR.<\/p>\n\n\n\n
          But sometimes this fails, especially when:<\/p>\n\n\n\n
            \n
          • the PR branch was deleted<\/li>\n\n\n\n
          • the PR is already merged<\/li>\n\n\n\n
          • the PR came from a fork<\/li>\n\n\n\n
          • the local remote is not configured correctly<\/li>\n\n\n\n
          • GitHub CLI tries to fetch a branch that no longer exists<\/li>\n<\/ul>\n\n\n\n
            That is why the script includes a fallback.<\/p>\n\n\n\n
            \n\n\n\n
            7. Fallback checkout using GitHub PR refs<\/h2>\n\n\n
            git fetch \"$REMOTE\"<\/span> \"+refs\/pull\/${PR}\/head:refs\/remotes\/${REMOTE}\/${BRANCH}\"<\/span>\ngit switch<\/span> -C \"$BRANCH\"<\/span> \"refs\/remotes\/${REMOTE}\/${BRANCH}\"<\/span>\n<\/code><\/span>Code language:<\/span> JavaScript<\/span> (<\/span>javascript<\/span>)<\/span><\/small><\/pre>\n\n\n
            This fetches the PR directly from GitHub\u2019s pull request reference.<\/p>\n\n\n\n
            For example, for PR 56<\/code>, it fetches:<\/p>\n\n\n
            refs\/pull\/56\/head\n<\/code><\/span><\/pre>\n\n\n
            This is often more reliable than fetching the original branch name.<\/p>\n\n\n\n
            This fallback fixed your earlier issue:<\/p>\n\n\n
            fatal<\/span>: couldn't find remote ref refs\/heads\/update_aws\n<\/code><\/span>Code language:<\/span> HTTP<\/span> (<\/span>http<\/span>)<\/span><\/small><\/pre>\n\n\n
            That happened because the branch update_aws<\/code> was no longer available as a normal branch.<\/p>\n\n\n\n
            \n\n\n\n
            8. Continue even if checkout fails<\/h2>\n\n\n\n
            If checkout still fails, the script does not stop immediately.<\/p>\n\n\n\n
            Instead, it continues with:<\/p>\n\n\n
            gh pr diff \"$PR\"<\/span> --patch\n<\/code><\/span>Code language:<\/span> JavaScript<\/span> (<\/span>javascript<\/span>)<\/span><\/small><\/pre>\n\n\n
            That means Claude can still review the pull request from the diff, even if the branch cannot be checked out locally.<\/p>\n\n\n\n
            This is useful for merged or closed PRs.<\/p>\n\n\n\n
            \n\n\n\n
            9. Collect PR context<\/h2>\n\n\n\n
            The script writes everything to:<\/p>\n\n\n
            \/tmp\/pr-56-context.md\n<\/code><\/span><\/pre>\n\n\n
            It includes:<\/p>\n\n\n\n
              \n
            • repository name<\/li>\n\n\n\n
            • PR state<\/li>\n\n\n\n
            • base branch<\/li>\n\n\n\n
            • head branch<\/li>\n\n\n\n
            • checkout status<\/li>\n\n\n\n
            • PR metadata<\/li>\n\n\n\n
            • changed files<\/li>\n\n\n\n
            • patch\/diff<\/li>\n<\/ul>\n\n\n\n
              This context is then sent to Claude.<\/p>\n\n\n\n
              \n\n\n\n
              10. Ask Claude to review<\/h2>\n\n\n\n
              The script pipes the context into Claude:<\/p>\n\n\n
              cat \"$CTX\"<\/span> | claude -p \"...\"<\/span>\n<\/code><\/span>Code language:<\/span> JavaScript<\/span> (<\/span>javascript<\/span>)<\/span><\/small><\/pre>\n\n\n
              The prompt tells Claude to focus on important review areas:<\/p>\n\n\n\n
                \n
              • correctness bugs<\/li>\n\n\n\n
              • security risks<\/li>\n\n\n\n
              • missing tests<\/li>\n\n\n\n
              • API compatibility issues<\/li>\n\n\n\n
              • race conditions<\/li>\n\n\n\n
              • edge cases<\/li>\n\n\n\n
              • maintainability issues that matter<\/li>\n<\/ul>\n\n\n\n
                It also tells Claude not to edit files.<\/p>\n\n\n\n
                That matters because for review workflows, we want feedback, not automatic code changes.<\/p>\n\n\n\n
                \n\n\n\n
                11. Save the review<\/h2>\n\n\n\n
                Claude\u2019s output is saved to:<\/p>\n\n\n
                claude-pr-56-review<\/span>.md<\/span>\n<\/code><\/span>Code language:<\/span> CSS<\/span> (<\/span>css<\/span>)<\/span><\/small><\/pre>\n\n\n
                This lets you review the AI output before posting.<\/p>\n\n\n\n
                \n\n\n\n
                12. Optionally post to GitHub<\/h2>\n\n\n\n
                If you run:<\/p>\n\n\n
                .\/review.sh 56 --post\n<\/code><\/span><\/pre>\n\n\n
                The script posts the review using:<\/p>\n\n\n
                gh pr comment \"$PR\"<\/span> --body-file \"$OUT\"<\/span>\n<\/code><\/span>Code language:<\/span> JavaScript<\/span> (<\/span>javascript<\/span>)<\/span><\/small><\/pre>\n\n\n
                For safer usage, I prefer:<\/p>\n\n\n
                .\/review.sh 56\nless claude-pr-56-review.md\ngh pr comment 56 --body-file claude-pr-56-review.md\n<\/code><\/span><\/pre>\n\n\n
                That keeps a human in the loop.<\/p>\n\n\n\n
                \n\n\n\n
                Real-world use cases<\/h1>\n\n\n\n
                Use case 1: Quick review before approving a PR<\/h2>\n\n\n\n
                A teammate opens a PR. Before you manually inspect it, run:<\/p>\n\n\n
                .\/review.sh 123\n<\/code><\/span><\/pre>\n\n\n
                Claude gives you a first-pass review. You can then focus your human review on the risky areas.<\/p>\n\n\n\n
                \n\n\n\n
                Use case 2: Reviewing infrastructure changes<\/h2>\n\n\n\n
                For Terraform, Kubernetes, Helm, or AWS modules, this is especially useful.<\/p>\n\n\n\n
                Claude can help spot:<\/p>\n\n\n\n
                  \n
                • overly broad IAM permissions<\/li>\n\n\n\n
                • missing encryption<\/li>\n\n\n\n
                • accidental public exposure<\/li>\n\n\n\n
                • risky security group changes<\/li>\n\n\n\n
                • missing variable validation<\/li>\n\n\n\n
                • breaking module interface changes<\/li>\n\n\n\n
                • missing README or example updates<\/li>\n<\/ul>\n\n\n\n
                  For Terraform repositories, you can customize the prompt with infrastructure-specific review goals.<\/p>\n\n\n\n
                  \n\n\n\n
                  Use case 3: Reviewing merged PRs<\/h2>\n\n\n\n
                  Sometimes you want to analyze an old PR after it has already been merged.<\/p>\n\n\n\n
                  Normal branch checkout may fail because the source branch was deleted.<\/p>\n\n\n\n
                  This script still tries to review from:<\/p>\n\n\n
                  gh pr diff\n<\/code><\/span><\/pre>\n\n\n
                  That makes it useful for audits, postmortems, and learning.<\/p>\n\n\n\n
                  \n\n\n\n
                  Use case 4: Preparing better code review comments<\/h2>\n\n\n\n
                  Instead of posting Claude\u2019s full output directly, you can use it as a draft.<\/p>\n\n\n\n
                  You read the generated Markdown, remove anything unnecessary, and post only the useful comments.<\/p>\n\n\n\n
                  This avoids noisy AI-generated reviews.<\/p>\n\n\n\n
                  \n\n\n\n
                  Suggested improvements<\/h1>\n\n\n\n
                  You can improve the script further by adding:<\/p>\n\n\n\n
                  Terraform-specific mode<\/h2>\n\n\n\n
                  Example:<\/p>\n\n\n
                  .\/review.sh 56 --terraform\n<\/code><\/span><\/pre>\n\n\n
                  Then use a Terraform-specific Claude prompt.<\/p>\n\n\n\n
                  Security-only mode<\/h2>\n\n\n\n
                  Example:<\/p>\n\n\n
                  .\/review.sh 56 --security\n<\/code><\/span><\/pre>\n\n\n
                  This would focus only on security and data exposure.<\/p>\n\n\n\n
                  Review without posting<\/h2>\n\n\n\n
                  Default behavior should stay safe:<\/p>\n\n\n
                  .\/review.sh 56\n<\/code><\/span><\/pre>\n\n\n
                  Auto-posting should remain explicit:<\/p>\n\n\n
                  .\/review.sh 56 --post\n<\/code><\/span><\/pre>\n\n\n
                  That is the right default.<\/p>\n\n\n\n
                  \n\n\n\n
                  Best practices<\/h1>\n\n\n\n
                  Do not blindly post AI-generated reviews.<\/p>\n\n\n\n
                  Use Claude as an assistant, not as the final reviewer.<\/p>\n\n\n\n
                  Prefer this flow:<\/p>\n\n\n
                  .\/review.sh 56\nless claude-pr-56-review.md\ngh pr comment 56 --body-file claude-pr-56-review.md\n<\/code><\/span><\/pre>\n\n\n
                  Keep prompts specific to the repository type.<\/p>\n\n\n\n
                  For example, a Terraform module review should focus on AWS security, IAM, networking, variables, outputs, examples, and provider compatibility.<\/p>\n\n\n\n
                  A backend service review should focus on API behavior, database changes, error handling, concurrency, and tests.<\/p>\n\n\n\n
                  A frontend review should focus on UI behavior, state management, accessibility, performance, and browser compatibility.<\/p>\n\n\n\n
                  \n\n\n\n
                  Final summary<\/h1>\n\n\n\n
                  This script gives you a lightweight AI-assisted PR review workflow using tools many developers already use:<\/p>\n\n\n
                  Git + GitHub CLI + Claude Code + Bash\n<\/code><\/span><\/pre>\n\n\n
                  It is fast, transparent, and easy to customize.<\/p>\n\n\n\n
                  The best part is that it keeps the human reviewer in control. Claude provides a first-pass review, but you decide what gets posted to the PR.<\/p>\n\n\n\n
                  That is exactly how AI code review should work: helpful, practical, and human-supervised.<\/p>\n","protected":false},"excerpt":{"rendered":"
                  How to Review GitHub Pull Requests from the Terminal Using gh and Claude Code Introduction Code review is one of the most important parts of software development,… <\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_joinchat":[],"footnotes":""},"categories":[11138],"tags":[],"class_list":["post-76154","post","type-post","status-publish","format-standard","hentry","category-best-tools"],"_links":{"self":[{"href":"https:\/\/www.devopsschool.com\/blog\/wp-json\/wp\/v2\/posts\/76154","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.devopsschool.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.devopsschool.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.devopsschool.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.devopsschool.com\/blog\/wp-json\/wp\/v2\/comments?post=76154"}],"version-history":[{"count":1,"href":"https:\/\/www.devopsschool.com\/blog\/wp-json\/wp\/v2\/posts\/76154\/revisions"}],"predecessor-version":[{"id":76155,"href":"https:\/\/www.devopsschool.com\/blog\/wp-json\/wp\/v2\/posts\/76154\/revisions\/76155"}],"wp:attachment":[{"href":"https:\/\/www.devopsschool.com\/blog\/wp-json\/wp\/v2\/media?parent=76154"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.devopsschool.com\/blog\/wp-json\/wp\/v2\/categories?post=76154"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.devopsschool.com\/blog\/wp-json\/wp\/v2\/tags?post=76154"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}