web-test
Warn
Audited by Gen Agent Trust Hub on Jun 24, 2026
Risk Level: MEDIUMREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill uses
AsyncFunctioninscripts/run.mjsto execute JavaScript code generated at runtime. In interactive mode, it also starts a local HTTP server on127.0.0.1that exposes an unauthenticated/execendpoint, allowing arbitrary code execution. - [DATA_EXFILTRATION]: The runtime sandbox in
scripts/run.mjsprovides access to thefs.readFileSyncandfs.writeFileSyncfunctions. When combined with Playwright's network capabilities, this enables the potential reading and exfiltration of sensitive local data. - [COMMAND_EXECUTION]: The
scripts/browser.mjsfile executes external binaries, specificallyffmpeg, usingchild_process.spawnandchild_process.execFileSyncto support the video recording feature.
Audit Metadata