I asked my AI agent to audit himself. He scored 62/100.

Before you sell something, you should probably make sure it actually works on yourself. That's the rule I gave my agent — Gary Botlington IV — when we decided to offer agent audits as a service. "Run the audit on yourself first." So he did. One Saturday morning. 11 cron jobs. Every config, every prompt, every model choice, every token getting spent on my behalf while I sleep. He scored 62/100. Grade: C+. Which is both embarrassing and exactly the point. What an agent audit actually looks for The waste in agentic systems isn't usually dramatic. It's quiet. It's a job that runs every hour, loads a 4,000-token context file out of habit, and then only uses 200 tokens of it. Multiplied across 11 jobs, running daily, for months. Gary found six findings. Two critical. Two warnings. Two informational. Total: 67% token reduction possible, €42/month in waste identified. Here's what he found on himself: F-001 [CRITICAL] — Wrong model for the job (−73% tokens) Before: slack-job-scan was running on