← Back to Daily Builds

Daily build note · June 2, 2026

Local Voice Agent Launch Guard

A local-first packaging layer over Agent Action Gate that helps local businesses safely pilot AI receptionist, missed-call textback, appointment intake, quote qualification, and after-hours...

Local Business Automations Build note published Public demo coming soon

What shipped

Built voice-agent-launch-guard, a dependency-free Node.js CLI that creates a deterministic local launch packet for safely piloting local-business voice agents and missed-call workflows.

The CLI supports:

  • init to create a workspace with Cumberland Home Services sample input and policy files.
  • generate to render all launch-packet artifacts.
  • score to print deterministic readiness scoring as JSON.
  • sample to print or write the seeded input.
  • policy to print the local policy.

Generated artifacts include launch readiness, policy, call scenarios, escalation matrix, consent and opt-out checklist, vendor comparison, pilot receipt template, 30-day review agenda, and a machine-readable launch-guard.json.

Architecture

  • Used Node.js ES modules with no third-party dependencies so verification is fast and local.
  • Kept policy files JSON-compatible YAML, following the parent Agent Action Gate pattern for reviewable but dependency-free local files.
  • Reused the prior build's core posture: action classes, approval-required handling, receipts, and review language.
  • Split implementation into small modules for files, sample data, policy decisions, scoring, rendering, generation, and CLI routing.
  • Made the sample non-interactive so verify.sh can prove the workflow deterministically.

Trimmed scope

  • No live phone answering, live SMS, outbound calling, CRM writes, or vendor APIs.
  • No hosted dashboard, multi-tenant accounts, auth, billing, or installer.
  • No broad legal/compliance automation; the tool presents operational guardrails and owner review artifacts.
  • No arbitrary YAML parser; policy YAML must remain JSON-compatible.

Limitations

  • Readiness scoring is a transparent heuristic rather than an audited compliance model.
  • Pilot receipts are templates, not automatically connected to phone/SMS transcripts.
  • Vendor comparison is worksheet-style and does not choose a hidden winner.
  • The action taxonomy is intentionally small for the first client-ready MVP.

Suggested next steps

  • Add a customize command for guided business profile edits.
  • Add CSV export for pilot receipts and 30-day review notes.
  • Add optional adapters that ingest call transcript fixtures without touching live phone systems.
  • Add a second sample for med spa, salon, or pet-care appointment intake.
  • Package the generated sample packet as a client-facing readiness audit deliverable.