Context-Inappropriate Capability
Medium
- Confidence
- 88% confidence
- Finding
- The skill instructs the agent to perform PubMed or web searches and make external editorial decisions based on network-fetched content, but this capability is not declared in the manifest. Undeclared network access expands the skill's trust boundary, can leak user-provided manuscript content or references to third parties, and makes behavior harder to audit or permission-gate.
