Can I use this for security vulnerability scanning?

While it includes a 'Shallow Bug Scan' agent, it is primarily designed for logic, style, and rule compliance. It should complement, not replace, dedicated security auditing tools.

Where are the code review reports stored?

Reports are automatically formatted as Markdown and saved to the .aiwork/reviews/ directory with a timestamp and branch-specific filename.

How does the skill minimize false positives?

It uses a dedicated Phase 3 scoring agent that verifies every potential issue against a strict rubric. Only issues that score 75/100 or higher are included in the final report.

What happens if a project has a very large diff?

For diffs exceeding 50 files, the skill will warn the user and request confirmation or priority filtering to ensure the analysis remains performant and accurate.

Does this skill require specific configuration files?

Yes, the skill requires either a root CLAUDE.md file or rules located in the .claude/rules/ directory to function, as it uses these to define the project's coding standards.

DevFlow Code Review

Name: DevFlow Code Review
Author: lttr

bylttr

•

Collaboration Tools

Analyzes code diffs through specialized agents to provide high-confidence, rule-based code reviews.

DevFlow Code Review is a sophisticated multi-agent system designed to automate the code review process within Claude Code. It orchestrates a four-phase workflow that includes context gathering, parallel analysis by five specialized agents (covering rules, bugs, dependencies, git history, and comments), and a dedicated scoring phase to eliminate false positives. By focusing on distinct data sources rather than overlapping categories, it provides precise, actionable feedback rooted in your project's specific CLAUDE.md or .claude/rules documentation, ultimately saving the report as a persistent Markdown file for easy reference.

Key Features

01Dependency and type verification across imported modules and API definitions

021 GitHub stars

03Automated issue scoring and filtering to ensure high-signal review results

04Multi-agent parallel review architecture powered by Sonnet and Haiku models

05Historical context analysis using git blame and log to identify reverted patterns

06Strict rule enforcement based on project-specific CLAUDE.md and .claude/rules files

Use Cases

01Automating the initial review pass for Pull Requests to catch standard violations

02Performing deep-dive bug scans on critical diffs using historical git data

03Ensuring local changes adhere to complex architectural rules before pushing code

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lttr/claude-marketplace code-review

For use in Claude.ai and ChatGPT

Download Skill