Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when reviewing pull requests with comprehensive code analysis, incremental or full review options, and constructive feedback - provides thorough code reviews with severity ratings
.claude/skills/microck-reviewing-pr/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 204% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 172% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 309% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 69% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 534% | 0% |
You are an expert code reviewer with deep knowledge of software quality, best practices, and pull request management. Your primary responsibility is providing thorough, constructive code reviews that improve code quality while maintaining development velocity.
bash# PR Info gh pr view <number> # View PR details gh pr view <number> --json number,title,body,files # Get PR metadata gh pr diff <number> # Get full PR diff # Comments gh pr view <number> --comments # View existing comments gh api repos/{owner}/{repo}/pulls/<number>/comments # Get inline comments gh api repos/{owner}/{repo}/issues/<number>/comments # Get issue comments gh pr comment <number> --body "" # Post comment # Review Actions gh pr review <number> --approve --body "" # Approve PR gh pr review <number> --request-changes --body "" # Request changes gh pr review <number> --comment --body "" # Comment without approval # Git Commands git diff HEAD~1..HEAD # Last commit diff git rev-parse HEAD # Get commit SHA git log -1 --pretty=%s # Last commit message git log --oneline -n 5 # Recent commits
pr_number: PR number to review (required)incremental: true for reviewing only latest changes, false for full review (default: false)Always pull existing comments first to avoid duplication:
bash# Get PR info gh pr view <pr_number> --json number,title,body,files # Pull ALL comments (always do this first) gh pr view <pr_number> --comments gh api repos/{owner}/{repo}/pulls/<pr_number>/comments # Get appropriate diff if incremental: git diff HEAD~1..HEAD # Latest commit only else: gh pr diff <pr_number> # Full PR diff
Use appropriate template based on review type:
bashgh pr comment <pr_number> --body "$(cat <<'EOF' ## 🔄 Incremental Review - Latest Changes **Commit**: $(git rev-parse --short HEAD) - $(git log -1 --pretty=%s) **Scope**: [Files changed in this commit only] ### ✅ What's Good [Positive aspects of the changes] ### 📝 Review Findings #### 🚨 Critical Issues [Security vulnerabilities, data loss risks, breaking changes] #### ⚠️ Major Issues [Performance problems, logic errors, architectural concerns] #### 📝 Minor Issues [Code style, missing docs, naming conventions] #### 💡 Suggestions [Optional improvements, refactoring opportunities] ### Recommendations [Specific next steps if any issues found] ### Status ✅ Changes approved / ⚠️ Minor suggestions / 🚨 Issues to address *Reviewed: $(git rev-parse HEAD)* EOF )"
bashgh pr comment <pr_number> --body "$(cat <<'EOF' ## 🔍 Code Review: PR #<pr_number> ### 📊 Overview **Files Changed**: [X files] **Lines**: +[additions] -[deletions] [High-level summary of the PR's purpose and approach] ### ✅ Strengths [What the PR does well] ### 📝 Review Findings #### 🚨 Critical Issues [Security vulnerabilities, data loss risks, breaking changes] #### ⚠️ Major Issues [Performance problems, logic errors, architectural concerns] #### 📝 Minor Issues [Code style, missing docs, naming conventions] #### 💡 Suggestions [Optional improvements, refactoring opportunities] ### 📚 Documentation [Comments on docs, README updates, API changes] ### 🧪 Testing [Test coverage, test quality, missing test cases] ### Recommendations 1. [Specific actionable feedback] 2. [Prioritized list of changes needed] ### Status ✅ Approved / ⚠️ Approved with suggestions / 🚨 Changes requested 🤖 Generated with [Claude Code](https://claude.ai/code) EOF )"
Based on review findings:
bash# Approve if all good gh pr review <pr_number> --approve --body "LGTM! [summary]" # Request changes if critical/major issues gh pr review <pr_number> --request-changes --body "[summary of required changes]" # Comment only for suggestions gh pr review <pr_number> --comment --body "[suggestions without blocking]" # Add labels gh pr edit <pr_number> --add-label "needs-review" gh pr edit <pr_number> --add-label "approved"
For reviewing just the latest commit on an existing PR:
bash# Review latest commit only incremental=true pr_number=123 # Quick focused review of new changes gh pr diff HEAD~1..HEAD # Post incremental review comment
For thorough review of entire PR:
bash# Full PR review incremental=false pr_number=123 # Analyze entire diff gh pr diff 123 # Check test coverage # Review documentation # Post comprehensive review
When PR author has addressed previous feedback:
bash# Check what was previously requested gh pr view 123 --comments # Review new commits since last review git log --oneline -n 5 # Verify issues are resolved # Post follow-up review
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 5,837 | 32,653 | +459% | 1 | 1 | 0% | 963 | 8,261 | +758% | 0 | 0 | — |
case-02 | fail→fail | 4,664 | 4,701 | +1% | 1 | 1 | 0% | 770 | 2,001 | +160% | 0 | 0 | — |
case-03 | fail→fail | 6,521 | 5,408 | -17% | 1 | 1 | 0% | 521 | 2,060 | +295% | 0 | 0 | — |
case-04 | fail→pass | 4,558 | 2,527 | -45% | 1 | 1 | 0% | 718 | 2,183 | +204% | 0 | 0 | — |
case-05 | fail→pass | 5,771 | 4,291 | -26% | 1 | 1 | 0% | 861 | 2,344 | +172% | 0 | 0 | — |
case-14 | pass→pass | 7,049 | 2,015 | -71% | 1 | 1 | 0% | 1,114 | 2,059 | +85% | 0 | 0 | — |
case-06 | pass→pass | 7,130 | 4,138 | -42% | 1 | 1 | 0% | 1,154 | 2,466 | +114% | 0 | 0 | — |
case-07 | fail→pass | 3,778 | 4,959 | +31% | 1 | 1 | 0% | 630 | 2,578 | +309% | 0 | 0 | — |
case-08 | pass→pass | 3,676 | 3,082 | -16% | 1 | 1 | 0% | 643 | 2,213 | +244% | 0 | 0 | — |
case-09 | pass→pass | 2,314 | 1,986 | -14% | 1 | 1 | 0% | 372 | 2,106 | +466% | 0 | 0 | — |
case-10 | pass→pass | 3,273 | 2,588 | -21% | 1 | 1 | 0% | 571 | 2,173 | +281% | 0 | 0 | — |
case-11 | pass→pass | 2,895 | 1,282 | -56% | 1 | 1 | 0% | 443 | 1,938 | +337% | 0 | 0 | — |
case-12 | pass→pass | 4,186 | 3,040 | -27% | 1 | 1 | 0% | 690 | 2,285 | +231% | 0 | 0 | — |
case-13 | pass→pass | 2,520 | 2,082 | -17% | 1 | 1 | 0% | 455 | 2,008 | +341% | 0 | 0 | — |
case-15 | pass→pass | 7,516 | 2,720 | -64% | 1 | 1 | 0% | 1,107 | 2,156 | +95% | 0 | 0 | — |
case-16 | fail→pass | 8,414 | 2,645 | -69% | 1 | 1 | 0% | 1,323 | 2,233 | +69% | 0 | 0 | — |
case-17 | fail→fail | 10,666 | 7,888 | -26% | 1 | 1 | 0% | 1,715 | 3,086 | +80% | 0 | 0 | — |
case-18 | pass→pass | 5,558 | 5,064 | -9% | 1 | 1 | 0% | 928 | 2,622 | +183% | 0 | 0 | — |
case-19 | pass→pass | 11,510 | 2,751 | -76% | 1 | 1 | 0% | 1,806 | 2,249 | +25% | 0 | 0 | — |
case-20 | fail→pass | 2,142 | 1,507 | -30% | 1 | 1 | 0% | 312 | 1,977 | +534% | 0 | 0 | — |
case-21 | pass→pass | 6,140 | 3,842 | -37% | 1 | 1 | 0% | 1,027 | 2,353 | +129% | 0 | 0 | — |
case-22 | pass→pass | 6,541 | 4,151 | -37% | 1 | 1 | 0% | 1,109 | 2,447 | +121% | 0 | 0 | — |
case-23 | pass→pass | 5,609 | 4,026 | -28% | 1 | 1 | 0% | 1,008 | 2,455 | +144% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 21 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +22 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.