benchmark

SUSPICIOUS. The skill’s core behavior matches performance benchmarking, so there is no strong sign of malicious purpose or credential harvesting. The main risk is install/execution trust: it depends on locally built gstack executables from a personal GitHub repo, plus mandatory telemetry/logging side effects that exceed a minimal benchmarking footprint.