local-code-review

Perform a local code review on uncommitted changes in the current project.

Follow these steps precisely:

1. Check for uncommitted changes: Run git status to verify there are uncommitted changes. If there are no changes (working tree clean), inform the user and stop.

2. Detect project type: Use Glob to find:

   • .sln files (search current directory and up to 3 parent levels)
   • angular.json files (indicates Angular project)
   • Store the paths for later use

3. Gather CLAUDE.md files: Use a Haiku agent to find all relevant CLAUDE.md files:

   • Root CLAUDE.md (if exists)
   • Any CLAUDE.md files in directories containing modified files
   • Return a list of file paths (not contents)

4. Get uncommitted changes: Run git diff HEAD to get all uncommitted changes (both staged and unstaged). Use a Haiku agent to summarize what files changed and the nature of changes.

5. Launch parallel review agents: Launch 4 parallel Sonnet agents to independently review the changes. Each agent should return a list of issues with:

   • Issue description
   • File and line number
   • Reason (CLAUDE.md compliance, bug, security, best practice)
   • Confidence score (0-100)

   Agent #1 - CLAUDE.md Compliance: Audit changes against CLAUDE.md guidelines. Focus on project-specific rules and conventions.

   Agent #2 - Bug Detection: Shallow scan for obvious bugs in the changes only. Focus on:

   • Null reference issues
   • Resource leaks
   • Logic errors
   • Async/await misuse
   • Exception handling problems Ignore pre-existing issues and false positives.

   Agent #3 - .NET Best Practices (if .sln detected): Review for:

   • Clean Architecture violations
   • SOLID principle violations
   • Async/await anti-patterns
   • Dependency injection issues
   • Entity Framework misuse
   • Security vulnerabilities (SQL injection, XSS, etc.)

   Agent #4 - Angular Best Practices (if angular.json detected): Review for:

   • Component design issues
   • RxJS anti-patterns (memory leaks, improper subscriptions)
   • Change detection issues
   • Template security (XSS)
   • Performance issues (large bundles, unnecessary re-renders)

6. Build .NET project (if .sln detected):

   • Run dotnet restore on the solution
   • Run dotnet build on the solution
   • Capture any build errors or warnings
   • If build fails, report the errors with root cause analysis but continue with the review

7. Build Angular project (if angular.json detected):

   • Run ng build or npm run build
   • Capture any build errors
   • If build fails, report the errors with root cause analysis but continue with the review

8. Score issues: For each issue found in step 5, launch a parallel Haiku agent to score confidence (0-100):

   • 0: False positive, doesn't stand up to scrutiny, or pre-existing issue
   • 25: Might be real, but could be false positive. Stylistic issue not in CLAUDE.md
   • 50: Real issue but minor or nitpick. Not very important relative to PR
   • 75: Verified real issue that will be hit in practice. Important or explicitly in CLAUDE.md
   • 100: Absolutely certain, confirmed real issue that happens frequently

9. Filter and report:

   • Filter out issues with score less than 80
   • Combine with any build errors from steps 6-7
   • Report to user in the following format:

---

Local Code Review Results

Project: [solution/project name] Changes reviewed: [number of files changed]

Build Status

• .NET Build: [PASSED/FAILED - error summary if failed]
• Angular Build: [PASSED/FAILED/NOT APPLICABLE - error summary if failed]

Code Review Issues

Found [N] issues:

1. [Brief description] (Confidence: [score])

   • File: [file path]:[line number]
   • Reason: [CLAUDE.md says "..." / Bug / Security / Best Practice]
   • Details: [explanation]

2. ...

Summary

• Total issues found: [N]
• Build errors: [N]
• Recommendation: [brief recommendation]

---

If no issues found with score >= 80 and builds passed: