Skill-Inject: Measuring Agent Vulnerability to Skill File Attacks

There's nothing to measure, prompt injection is unsolvable