Vague Triggers
Medium
- Confidence
- 82% confidence
- Finding
- The trigger list includes broad natural-language phrases like 'Is this video real?' and 'deepfake detection' that can match ordinary user requests outside a narrowly scoped KYC workflow. If the agent auto-selects this skill from such phrases, it may send sensitive user media to a remote endpoint unexpectedly, creating privacy and consent risks even though the skill states it is for consented media.
