Hey WondTech readers! Ever wonder what's really inside your AI agent's “brain” – its collection of skills? A new tool, skill-auditor v0.1.0, is here to peek under the hood and make sure everything is in order. This command-line tool acts like a meticulous detective, scanning your AI agent's abilities for all sorts of hidden problems. What does this mean for you, especially if you're working with AI agents? Well, as AI systems like Hermes, Claude Code, or OpenCode grow, their skill collections can balloon to hundreds of entries. These aren't just simple lists; they're practically executable code that handles commands and sometimes even sensitive credentials. Yet, these critical collections often go unchecked. Imagine shipping software with a huge codebase that no one has properly audited! This new auditor aims to fix that. It's designed to be agent-agnostic, meaning it can work with various AI systems, starting with Hermes. It looks for common but dangerous issues. For instance, it can spot structural problems like duplicate folders causing conflicts, broken references from typos or missing files, and un-delivered scripts that your AI agent might be trying to call. In a real-world test with 200 Hermes skills, a direct audit found a nested duplicate folder leading to runtime errors, a broken reference due to a simple accent typo, and several missing script links. Crucially, skill-auditor also checks for hardcoded secrets – things like API keys for AWS, GitHub, OpenAI, or Google. While the initial audit found secrets hygiene to be clean (mostly documentation placeholders), relying on a one-time check isn't a robust process. This tool makes secret scanning repeatable and systematic. Ultimately, skill-auditor v0.1.0 provides a clear classification for each skill, telling you if it's OK, a WARNING, or BROKEN. This helps maintain the integrity and security of your AI agents' capabilities, ensuring they run smoothly and securely.