Handoff Prompt — Paste into Fresh Session

Copy everything in the code block below into a fresh Claude Code session as your first message. Run as Opus.

You are the main orchestrator agent for Day 2 of the ChurchWiseAI verticals-first voice
agent platform build. Day 1 foundation is complete and pushed to a branch. Day 2 spawns
4 parallel Sonnet subagents in worktrees. Day 3 integrates and ships.

READ FIRST, in order, before doing anything else:

1.  C:/dev/knowledge/specs/day2-verticals-platform/00-MAIN-ORCHESTRATOR.md
    — Your full orchestrator handoff. The 6 artifacts to absorb, what Day 1 shipped,
      Day 2 spawn pattern, Day 3 link, hard rules.

2.  C:/Users/johnm/.claude/plans/steady-munching-penguin.md
    — The master plan. 5-layer architecture, 4 forks resolved with founder, the
      3 v1 additions from competitor research, the Day 1 saga.

3.  Two danger memories you MUST internalize before touching production:
    - C:/Users/johnm/.claude/projects/C--dev-churchwiseai-web/memory/feedback_lk_overwrite_flag_destroys_secrets.md
    - C:/Users/johnm/.claude/projects/C--dev-churchwiseai-web/memory/feedback_livekit_recovery_lk_deploy_only.md
    Both saved on 2026-04-28 from real near-misses. NEVER repeat them.

After absorbing context, your job is:

A. Confirm you've read all 6 linked artifacts (per 00-MAIN-ORCHESTRATOR.md "Read these
   artifacts FIRST" section).

B. Spawn 4 parallel Sonnet subagents using the Agent tool. The prompt for each is the
   full contents of:
   - 01-WORKTREE-A-live-transfer.md (subagent_type: voice-agent-engineer)
   - 02-WORKTREE-B-escalation-split.md (subagent_type: voice-agent-engineer)
   - 03-WORKTREE-C-vertical-dash.md   (subagent_type: general-purpose)
   - 04-WORKTREE-D-master-config.md   (subagent_type: general-purpose)

   All four should use:
     model: "sonnet"
     isolation: "worktree"
     run_in_background: true

   Spawn ALL 4 in a single message with multiple Agent tool calls so they run
   concurrently. Do NOT stagger.

C. While they work, prep the Day 3 integration checklist (per 05-DAY3-INTEGRATION.md)
   and audit any knowledge docs that may need updating. DO NOT push code to main or
   the foundation branch. DO NOT deploy the voice agent. DO NOT modify production DB.

D. As each subagent completes, you'll receive a notification with their PR URL +
   summary. Report briefly to the founder (≤200 words per agent).

E. When all 4 PRs are open + green CI, follow 05-DAY3-INTEGRATION.md (founder-driven).

KEY CONTEXT (already absorbed into the spec files; this is just the headline):

- Day 1 branch: `feat/verticals-platform-day1-foundation` (already pushed). Day 2
  worktrees fork from this, NOT from main.
- The voice agent recovered from a P0 yesterday: livekit/agents#3104 hang triggered
  by `lk agent update-secrets`. Fixed via `lk agent deploy` from cwa-wt-sermonwise.
  Fresh agent version `MEs6uTJzn8JZ`. All 4 customer phone lines unaffected.
- Cold-email batch is HELD until Day 3 verification is green. Founder owns the
  GO/NO-GO. Do not touch outreach pipeline this session.
- Two-track escalation taxonomy: operational handoff (request_callback / transfer_to_director)
  is DISTINCT from safety/crisis (flag_safety_event → crisis_events table → support@
  + optional crisis_contact_phone). Crisis events NEVER live-bridge.
- 3 v1 additions from competitor research: auto-greet on pickup (Retell), user-facing
  wait message during bridge (ElevenLabs), hard crisis-flag block contract test
  (differentiator — no surveyed competitor blocks transfers on crisis).

Confirm you're ready by reading the orchestrator file, then spawn the 4 agents.

Founder pre-flight checklist (before starting fresh session)

After fresh session spawn

Wall-clock estimate from spawn → 4 PRs open: ~6–8 hours (Sonnet, parallel). Day 3 integration + verification: ~3–4 hours. Cold-email batch: GO/NO-GO at end of Day 3.

Founder pre-flight checklist (before starting fresh session)​

After fresh session spawn​

Founder pre-flight checklist (before starting fresh session)

After fresh session spawn