Context-Inappropriate Capability
Medium
- Confidence
- 83% confidence
- Finding
- The script delegates searching to an external CLI (openclaw) and passes user-controlled topic text into that tool without any validation or output trust controls. Even though the subprocess call itself is not shell-injectable, this broadens the attack surface to whatever parsing, plugin behavior, or prompt-handling the external tool performs, making this more dangerous in an agent skill that may run automatically on schedules.
