How does this skill reduce code review noise?

It requires Claude to verify the execution path and check if an issue is handled elsewhere (like in middleware or framework logic) before it is allowed to create a finding.

How does it handle test or generated code?

It applies different standards to test code (allowing hardcoded values) and ignores generated code unless specifically hand-edited, ensuring reviews stay relevant.

Will it flag intentional deviations in the codebase?

No, the skill instructs Claude to prioritize consistency with existing codebase patterns over isolated 'best practice' improvements that might violate local conventions.

Can it be used for security audits?

Absolutely. It ensures security findings are valid by checking for framework-level sanitization, parameterized queries, and API-layer validation before flagging a vulnerability.

Does it support specific frameworks?

Yes, it is designed to recognize framework-specific conventions in React, ORMs, and dependency injection systems where standard rules might not apply.

False Positive Prevention

Name: False Positive Prevention
Author: bitwarden

bybitwarden

•

Herramientas de Colaboración

Reduces noise in AI-driven code reviews by verifying execution paths and framework semantics before flagging potential issues.

This skill serves as a quality control layer for AI-driven code reviews, ensuring that Claude only flags genuine bugs rather than intentional patterns or framework-specific behaviors. It mandates a rigorous three-step verification process—tracing execution paths, checking external handlers, and confirming framework API contracts—to prevent the creation of low-value or incorrect findings. By respecting existing codebase conventions and recognizing intentional simplicity, this skill makes automated code reviews more actionable and significantly less intrusive for development teams.

Características Principales

011 GitHub stars

02Codebase convention alignment to prioritize consistency over isolated improvements

03Framework-aware analysis for React, dependency injection, and ORMs

04Three-step verification process for every code review finding

05Automated identification of generated code and test suites

06Detection of external handlers such as middleware, error boundaries, and validators

Casos de Uso

01Reviewing legacy codebases while maintaining existing architectural and stylistic patterns

02Conducting high-precision security audits where false positives waste developer time

03Integrating AI code reviews into CI/CD pipelines to ensure only critical issues are reported

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add bitwarden/ai-plugins avoiding-false-positives

For use in Claude.ai and ChatGPT

Download Skill