Can it integrate with my task management system?

The skill detects work item tracking systems (like GitHub Issues or Azure DevOps) and can automatically defer complex findings that require human design decisions.

How does it handle context window limits?

It utilizes subagents for the heavy lifting of reviewing and fixing, which keeps the main conversation context lean and allows for multiple iterations without hitting token limits.

Does it support Test-Driven Development (TDD)?

Yes. For behavioral bugs, the skill can write a failing test, implement the fix, and verify the green state automatically using your local test infrastructure.

How many times does the auto-review loop iterate?

By default, the skill runs up to 3 iterations to ensure code stability, but this can be adjusted using the --max-iterations flag.

Auto Review & Fix Loop

Name: Auto Review & Fix Loop
Author: alphaleonis

byalphaleonis

0•

Security & Testing

Automates the end-to-end code review process by identifying issues, triaging findings, and autonomously applying fixes until the code stabilizes.

This skill provides an autonomous agentic workflow for continuous code quality. It executes a cyclical loop of reviewing code using a subagent, triaging findings based on severity and confidence, and applying fixes—optionally using Test-Driven Development (TDD) patterns. By isolating context in subagents and intelligently deciding whether to re-review based on change magnitude, it ensures that even complex refactors or large sets of uncommitted changes are polished and regression-free before human intervention is required.

Key Features

01Subagent context isolation to prevent context window exhaustion during long tasks

02Intelligent triage based on severity, confidence scores, and issue patterns

03Automatic deferral of complex design issues to work tracking systems like GitHub Issues or Azure DevOps

040 GitHub stars

05Integrated TDD workflow that writes failing tests before applying behavioral fixes

06Autonomous review-fix-recheck loop that iterates until code stability

Use Cases

01Cleaning up a large batch of uncommitted changes before submitting a pull request

02Automatically identifying and fixing security vulnerabilities across a repository

03Ensuring new features adhere to project conventions through iterative re-reviews and regression checks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add alphaleonis/decaf-claude-config auto-review

For use in Claude.ai and ChatGPT

Key Features

01Subagent context isolation to prevent context window exhaustion during long tasks

02Intelligent triage based on severity, confidence scores, and issue patterns

03Automatic deferral of complex design issues to work tracking systems like GitHub Issues or Azure DevOps

040 GitHub stars

05Integrated TDD workflow that writes failing tests before applying behavioral fixes

06Autonomous review-fix-recheck loop that iterates until code stability

Use Cases

01Cleaning up a large batch of uncommitted changes before submitting a pull request

02Automatically identifying and fixing security vulnerabilities across a repository

03Ensuring new features adhere to project conventions through iterative re-reviews and regression checks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add alphaleonis/decaf-claude-config auto-review

For use in Claude.ai and ChatGPT