Description-Behavior Mismatch
Medium
- Confidence
- 93% confidence
- Finding
- The documentation instructs users to use curl to download arbitrary image URLs directly to the filesystem, which exceeds the declared screenshot/browser-capture scope and introduces an uncontrolled network-to-disk write path. In an agent setting, this can be abused to fetch unexpected content, store untrusted files locally, or bypass any safety expectations tied to the browser toolchain.
