name: open-code-review-scan description: AI code quality gate that detects hallucinated imports, phantom packages, stale APIs, and AI-specific code defects. Use when the user asks to scan code, review code quality, check for AI-generated bugs, audit imports, find phantom dependencies, or run a code quality gate. Supports TypeScript, JavaScript, Python, Java, Go, Kotlin. license: MIT metadata: author: raye-deng version: "1.0" mcp-server: "@opencodereview/mcp-server"
Open Code Review — AI Code Quality Gate
Scan source code for AI-generated defects that traditional linters miss. This skill detects:
- Hallucinated imports: packages that don't exist or aren't in dependencies
- Phantom packages: code referencing non-existent modules
- Stale APIs: calls to removed or renamed APIs
- Dead code paths: unreachable code introduced by AI generation
- Version mismatches: code using APIs from wrong library versions
When to activate
- User asks to "scan code", "review code quality", "check for AI bugs"
- User wants to audit imports or find phantom dependencies
- User needs a code quality gate before merging PRs
- User mentions "code review", "lint", "static analysis" for AI-generated code
- User wants to check a diff or specific files for defects
Instructions
- Use the
scan_directorytool to scan the target codebase:- Provide the absolute path to the directory
- Optionally specify language filter (e.g., "typescript", "python")
- For PR reviews, use
scan_diffwith the diff content - Review the scan results for defect categories:
- Critical: hallucinated imports, missing dependencies
- Warning: stale APIs, version mismatches
- Info: dead code paths, unused variables
- For each defect, use
explain_issueto get detailed explanations and fix suggestions - To auto-fix issues, use
heal_codeon affected files - Present a structured report to the user with:
- Summary statistics (total issues, by severity)
- Top critical issues with explanations
- Recommended fixes
Supported Languages
TypeScript, JavaScript (Node.js), Python, Java, Go, Kotlin
Tips
- For best results, scan the project root so the tool can read
package.json/requirements.txt/pom.xmlfor dependency validation - Run
scan_diffon PR diffs for fastest feedback - Use
heal_codesparingly — always review auto-fixed code before committing
Related skills
TDD Red-Green-Refactor
Testing
Skill that guides Claude through the complete TDD cycle.
Claude CodeWindsurfintermediate
650
189
2,227
Web Accessibility Audit
Testing
Performs a comprehensive web accessibility audit following WCAG standards.
claudeCursorWindsurfintermediate
129
37
532
UAT Test Case Generator
Testing
Generates structured and comprehensive user acceptance test cases.
claudechatgptintermediate
87
24
454