computer-use

SUSPICIOUS: tujuan utamanya koheren untuk otomasi riset, dan penggunaan Orca sebagai primary/fallback UI masih masuk akal. Risiko utama berasal dari instruksi instalasi skill transitif pihak ketiga, termasuk repo personal, serta fallback vision AI yang dapat menerima credential model dan mengirim data ke endpoint yang dapat dikustomisasi.

This module is not overtly malicious by itself, but it functions as an execution wrapper that can trigger npx-based installation/registration of third-party “skills” based on a user-selected index. The key security concerns are (1) supply-chain risk from running external package manager commands without visible integrity/pinning controls and (2) subprocess.run(..., shell=True) combined with returning/printing stdout/stderr from the invoked process. Review the external artifacts’ trust/integrity controls and consider removing shell=True (use argv list) and minimizing/omitting stdout/stderr in outputs where appropriate.