personalized-outbound-ab-test

SUSPICIOUS: The skill is coherent as an outbound experimentation workflow and shows no supply-chain, credential, or exfiltration behavior in the provided text. However, it enables scaled autonomous messaging with limited per-action user approval and includes silent fallback behavior, which makes the operational risk materially higher than a read-only advisory skill.