Deployed agents skills

Design an AI appointment-setting agent (chat or voice) that books qualified prospects straight onto a calendar (persona, booking conversation flow, light qualification, real-time availability lookup, confirmation and reminder logic, and CRM capture) then emit a deploy spec. The skill designs and configures; the live agent runs on the owner's MCP plus a chat or voice provider.

Design a website chat agent that qualifies inbound visitors and books the qualified ones (persona, qualifying conversation flow, lead scoring, knowledge grounding, calendar booking, and CRM capture) then emit a deploy spec for a web chat widget. The skill designs and configures; the live widget runs on the owner's MCP plus a chat embed on the client site.

Drive a real computer or browser safely to automate a desktop/GUI task, our own governed operating layer over any computer-use runtime (native tools, a hosted API, or any desktop-control MCP), never a dependency on one.

Design an AI executive-assistant agent that manages the owner's calendar, inbox, and to-dos from a chat channel (Telegram/Slack/email) (reads the calendar and email, drafts replies in the owner's voice, schedules/reschedules events, builds and updates a to-do list, and surfaces a daily brief) then emits a deploy spec.

Deployed-agent category hub: routes to the right leaf for customer-facing agents (chat lead qualification, support, voice, appointment setter), personal agents (executive assistant, second brain), and the rails they run on (computer use, scheduling orchestration, payment rails). These design a deployable agent spec, not a live agent.

Safe agent transaction rails: issue a single-use, merchant- and amount-locked, time-boxed virtual card inside a HARD, human-set, revocable spend envelope; in-envelope micro-charges auto-authorize, above-threshold or off-allowlist spend routes to a human, every authorization is ledgered. Transacts WITHIN a pre-authorized envelope. Not autonomous spend, never money movement (G8).

Cross-party booking broker: merges free/busy across the agency, client, and CRM calendars (Google, Outlook, CalDAV), then books atomically via a two-phase commit (hold → confirm → release) so a slot is never double-booked. The atomic-booking infrastructure beneath the as-agent-appointment-setter conversation.

Scaffold a personal executive-assistant / second brain for Claude Code, a governed context layer (CLAUDE.md + @-imported context files + rule files + an append-only decision log + a backlog) that makes the assistant know you and get smarter without re-explaining.

Design an AI customer-support agent grounded in the client's own docs and knowledge base, persona, support conversation flow, retrieval grounding, ticket-deflection logic, and clean escalation to a human with full context. The skill designs and configures; the live agent runs on the owner's MCP plus a chat (or voice) channel.

Design, script, and configure an inbound or outbound AI voice agent and emit a provider-ready deploy spec, persona, conversation state machine, knowledge grounding, voice selection, guardrails, and escalation-to-human. The skill produces the design + config + deploy spec; the live agent runs on the owner's MCP plus a voice provider (Twilio/Vapi/Retell plus ElevenLabs).