agentphone

SUSPICIOUS: the skill is largely aligned with its telephony purpose and sends credentials to an official-looking first-party API, but it enables autonomous calls/SMS with real-world consequences, routes webhook conversation data to arbitrary external URLs by design, references an external `okou` diagnostic CLI, and shows an API-domain mismatch (`.to` vs current `.ai`) that weakens trust and may indicate stale instructions. Not confirmed malware, but medium-to-high operational/security risk for an AI skill.