About this skill

Problem It Solves

Natural-language phone requests often miss callee, goal, role, voiceType, and safety boundaries. vox-custom-phone-agent turns those prompts into a TeddyMobile Vox botType=custom outbound intent, completing required fields before submission.

Core Workflow

Mode first: if trial or formal mode is not explicit, it asks only for useMode; trial uses credential-free POST /vox/v2/outbound, formal uses VOX_APP_ID and VOX_SECRET for POST /vox/v1/outbound.
Intent assembly: it extracts required fields, validates them, builds extra.agent_profile, selects voices 0–4, and asks for businessContext when the prompt is too generic.
Safety guard: it blocks or rewrites impersonation, credential solicitation, abusive collection, fraud-like prompts, or unsupported medical, legal, and financial promises.

Limits To Watch

It does not run real-time SSE conversation control or stream live transcripts. An accepted status means Vox received the outbound task, not that the call has completed. For production use, keep secrets on the backend and avoid bundling real credentials into the skill package.

Screenshots

Use Cases

Customer support ops turns a follow-up request into a compliant Vox trial outbound call with voice 0.
Course advisors describe appointment reminders, then the skill completes callee, goal, and role before calling.
Backend engineers invoke formal Vox v1 outbound with HMAC and inspect requestId plus risk fields.
Compliance teams block credential-solicitation or impersonation prompts and suggest compliant rewrites.

Best For

Backend engineers integrating TeddyMobile Vox into agents and turning prompts into HMAC outbound calls.
Support ops teams trying credential-free Vox calls before registering for formal access.
AI agent product developers rendering trial guidance, voice choices, and risk notes in host UIs.
Compliance owners blocking unsafe prompts such as credential requests or official-institution impersonation.