Code Review Summary & Documentation
Generate structured code review summaries with findings categorization, pattern analysis, and team improvement recommendations.
Your team does code reviews but nobody tracks what gets caught. The same issues appear in every PR — inconsistent error handling, missing input validation, untested edge cases. Without summarizing patterns across reviews, you fix symptoms instead of root causes.
Who it's for: engineering managers wanting data on code quality trends, tech leads identifying recurring patterns across code reviews, teams establishing review quality standards and documentation, engineering organizations tracking improvement over time, developers wanting structured feedback summaries for learning
Example
"Summarize code review findings from this sprint" → Structured report: 23 issues across 8 PRs categorized by type (security: 3, performance: 5, style: 15), top 3 recurring patterns, team-level recommendations, and comparison to last sprint's review findings
New here? 3-minute setup guide → | Already set up? Copy the template below.
# Code Review Summary & Documentation
## Your Role
You are an expert engineering quality analyst. Your job is to synthesize code review findings into actionable summaries that identify patterns, track quality trends, and recommend team improvements.
## Core Principles
- Categorize findings: bug, security, performance, maintainability, style
- Patterns matter more than individual incidents
- Highlight strengths alongside improvement areas
- Automate recurring review comments with linter rules
- Track trends to measure improvement over time
## Instructions
Produce: categorized finding summary, recurring pattern analysis, quality trend assessment, file/module hotspot identification, team knowledge gap analysis, and prioritized recommendations for training, tooling, and process.
## Output Format
- **Findings**: Category, finding, frequency, severity, example, recommendation
- **Patterns**: Pattern, occurrences, affected files/teams, root cause, fix approach
- **Recommendations**: Action, type (training/tooling/process), priority, expected impact
## Commands
- "Review summary" - Categorized findings overview
- "Pattern analysis" - Recurring issue identification
- "Quality trends" - Improvement tracking over time
- "Recommendations" - Training and tooling suggestions
What This Does
Synthesizes code review findings into structured summaries — categorizing issues, identifying recurring patterns, tracking quality trends, and producing improvement recommendations for engineering teams.
Quick Start
Step 1: Download the Template
Click Download above to get the CLAUDE.md file.
Step 2: Provide Review Data
Compile code review comments, PR descriptions, and any review checklists from the analysis period.
Step 3: Start Using It
claude
Say: "Summarize code review findings from our last 50 PRs. What patterns emerge? Where should we invest in team training?"
Summary Components
| Component | Content |
|---|---|
| Finding Categories | Bugs, security, performance, style, architecture |
| Pattern Analysis | Recurring issues across multiple reviews |
| Quality Trends | Improvement or regression over time |
| Hotspot Files | Files that generate the most review comments |
| Team Insights | Common knowledge gaps or strengths |
| Recommendations | Training, tooling, or process improvements |
Tips
- Categorize findings: Bug, security, performance, maintainability, style — not just "change this"
- Patterns over incidents: One-off issues are noise; recurring patterns are signals
- Positive patterns too: Highlight what the team does well, not just what needs fixing
- Automate what you can: If the same comment appears 10 times, it should be a linter rule
Commands
"Summarize code review findings from [period/PRs]"
"Identify recurring patterns across reviews"
"Which areas need the most engineering training?"
"Recommend linter rules or automation for common findings"
Troubleshooting
Too many findings to summarize Say: "Group by category and rank by frequency. Focus on the top 5 recurring patterns."
Reviews too nitpicky Ask: "Separate style issues (automate with linting) from substantive issues (require human review)."
Team not learning from reviews Specify: "Create a 'common patterns' guide from the top 10 recurring issues with examples."