Does this skill automatically modify my agent files?

By default, it only provides a report. However, if you use the --fix option, the skill can automatically apply recommended improvements to your subagent files.

How do I run a review for a specific agent?

You can trigger a review by typing '/agent-reviewer' followed by the agent name or the specific file path of the subagent you wish to analyze.

Can I review all my subagents at once?

Yes, by using the --all flag, the skill will scan and provide a comprehensive review of every subagent file found in your .claude/agents/ directory.

What criteria does the Agent Reviewer use for evaluations?

It evaluates subagents based on single responsibility, tool scoping, prompt length, permission modes, and frontmatter accuracy to ensure they follow Claude's best practices.

Subagent Quality Reviewer

Name: Subagent Quality Reviewer
Author: thirdlf03

bythirdlf03

0•

安全与测试

Analyzes and optimizes Claude Code subagents based on industry best practices to ensure high-quality, reliable AI workflows.

The Subagent Quality Reviewer is a specialized auditing tool designed to maintain the performance and safety of your Claude Code ecosystem. It evaluates subagents against strict criteria such as single responsibility principles, proper tool scoping, and optimal prompt length. By generating detailed evaluation reports, scoring performance, and offering automated fixes, this skill helps developers standardize their custom AI agents, identify security risks in permission modes, and ensure that every subagent is perfectly tuned for its specific task.

主要功能

01Batch review capability to scan entire agent directories at once

02In-depth prompt length and output format auditing

03Automated frontmatter validation for subagent configurations

040 GitHub stars

05Best-practice analysis for tool scoping and permission settings

06Score-based evaluation reports with actionable improvement steps

使用场景

01Auditing a collection of custom subagents for security vulnerabilities and efficiency

02Standardizing agent configurations across large development teams and projects

03Troubleshooting underperforming agents by identifying prompt-overspecification

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add thirdlf03/claude-dev-toolkit agent-reviewer

For use in Claude.ai and ChatGPT

Download Skill