tao-run-on-lepton

SUSPICIOUS-LEANING but not malicious. The skill’s purpose and core capabilities mostly align with DGX Cloud Lepton job management, and credential flows target expected NVIDIA/S3 endpoints. Main concerns are install-trust ambiguity around the TAO package naming and the unpinned download-and-source of a remote GitHub shell script inside jobs, which raises medium supply-chain risk rather than clear malicious intent.