Context-Inappropriate Capability
Medium
- Confidence
- 88% confidence
- Finding
- The skill instructs the agent to fetch requirement documents directly from arbitrary URLs or invoke other skills to download them, which expands the trust boundary from local document processing to network retrieval. This can expose the agent to SSRF-like access patterns, unintended access to internal resources, or ingestion of untrusted remote content without explicit user confirmation and capability scoping.
