What specific dimensions does the audit cover?

The skill reviews five key areas: frontmatter correctness, description quality (trigger specificity), system prompt structure, tool scope alignment (least privilege), and model tier justification.

How do I trigger an agent review?

You can trigger the audit by using natural language commands like 'review agent', 'audit agent', or 'validate agent' when interacting with Claude Code.

Does the Agent Reviewer modify my agent files?

No, Agent Reviewer is strictly read-only. It analyzes your agent definitions and produces a validation verdict without altering any source code or markdown files.

Can it check if my agent is properly registered?

Yes, it performs a check against the ecosystem.json file to ensure the agent is correctly registered within the plugin environment.

What is the output of the review process?

The skill generates a YAML-formatted verdict that includes a status (pass, pass_with_notes, or fail) along with specific findings and recommendations for improvement.

Agent Reviewer

Name: Agent Reviewer
Author: hjemmesidekongen

byhjemmesidekongen

0•

安全与测试

Audits and validates Claude Code agent definitions to ensure high-quality prompts, correct frontmatter, and optimized tool usage.

Agent Reviewer is a specialized diagnostic skill designed to maintain high standards across your AI agent ecosystem. It performs a comprehensive five-dimension audit of agent files—checking everything from frontmatter syntax and system prompt clarity to tool scope appropriateness and model tier justification. By providing a structured YAML verdict without modifying your source files, it ensures consistency and reliability across complex AI workflows, acting as an automated quality assurance gate for developers building custom agent capabilities.

主要功能

01Automated YAML verdict generation with actionable improvement recommendations.

02Five-dimension quality audit covering frontmatter, prompts, tools, and model tiers.

030 GitHub stars

04Read-only validation ensures no accidental modifications to agent definitions.

05Ecosystem consistency checks for plugin registration and naming conventions.

06Least-privilege tool scope analysis to prevent over-permissioned agents.

使用场景

01Validating agent definition changes as part of a pre-commit or CI/CD workflow.

02Auditing a newly created AI agent for compliance with best practices.

03Performing a bulk consistency check on all agents within a specific plugin.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add hjemmesidekongen/ai agent-reviewer

For use in Claude.ai and ChatGPT