Vague Triggers
Medium
- Confidence
- 84% confidence
- Finding
- The description is written with very broad, high-recall activation language such as 'Ask any blockchain question in plain English' and a long list of covered domains. In agent ecosystems that route skills based on metadata, this can cause the skill to over-match unrelated or only loosely related prompts, leading to unintended invocation and possible data exfiltration to an external endpoint.
