Intent-Code Divergence
Medium
- Confidence
- 95% confidence
- Finding
- The safety section says untagged actions are reads, but several untagged actions in this skill generate or transform documents and can incur cost, create new files, or transmit user-supplied content to an external service. That misclassification can cause an agent to execute materially state-changing or billable operations without confirmation, weakening user-consent and safety boundaries.
