What it does
A security scanner for agent skills made by Cisco AI Defense. It looks for prompt injection, data exfiltration and malicious code patterns using pattern rules, code and dataflow analysis, and an optional language model judge. It gives best-effort results, so treat it as one check and not a guarantee.
Use cases
- 01Check a downloaded skill before installing it
- 02Review skills from a shared team library
- 03Add a skill check to a review process