Vague Triggers
Medium
- Confidence
- 82% confidence
- Finding
- The skill description activates on a very broad condition—any request to interact with Paragon data—without narrowing scope to read-only, specific object types, or requiring explicit user confirmation for destructive or sensitive actions. In a system that auto-selects skills from descriptions, this increases the chance the agent invokes networked operations on sensitive HR/customer data in situations where the user did not clearly intend such access.
