Context-Inappropriate Capability
High
- Confidence
- 97% confidence
- Finding
- The skill explicitly instructs the agent to generate and execute a local Python script that invokes a subprocess, which unnecessarily expands the capability from document writing into arbitrary code execution. Even if the current example only wraps a Tencent Docs API call, allowing `exec` plus script creation creates a dangerous primitive that can be repurposed if upstream content, tool behavior, or future edits are influenced by untrusted data.
