Customer Status Portal

You are building a Customer Status Portal (Block 14 in the Cafecito AI new-hire playbook). Full reference at https://cafecito-ai.com/new-hire/blocks/14-customer-status-portal. Read it. Use plan mode. Stop at every [GATE].

INPUTS YOU NEED FROM THE HUMAN (ask before doing anything else):
1. Prospect — sign shop, custom millwork, repair shop, property mgmt (Saul Signs and Signscenter are canonical first targets)
2. Decision-maker email
3. 5 sample job records — either real (anonymized) or synthetic. Each: customer name, project type, intake date, current stage, photo URLs if any, latest staff note

ENVIRONMENT (verify):
- Working dir: /home/eratner/cafecito-ai
- Cloudflare account: f7a9b24f679e1d3952921ee5e72e677e

SECRETS TO CONFIRM:
- ANTHROPIC_API_KEY (Claude — staff-note → customer-update translation)
- (R2 storage doesn't need a separate secret)

THE PLAN:

STEP 1 — RESEARCH + JOB PIPELINE STAGES (10 min)
- Pull prospect site. Identify their typical project flow (e.g. for signs: estimate → approval → fabrication → install).
- Define the canonical stages (4-6 max). Each gets a customer-friendly label and a staff label.
- Wait for the 5 sample job records.
[GATE 1 — show stages + 5 jobs parsed, ask "proceed?"]

STEP 2 — SCAFFOLD WORKER (15 min)
- Create /home/eratner/cafecito-ai/portal-<prospect-slug>/.
- Worker handles: GET /track/:id (customer view), GET /admin (staff view + add update), POST /api/job/:id/update (admin only — staff posts a note + Claude translates to customer-friendly).
- Bindings: D1 PORTAL_DB, R2 PORTAL_R2 (job photos).
- Routes: /portal-demo/<prospect-slug>/track/:id and /portal-demo/<prospect-slug>/admin.
[GATE 2 — show worker.js + D1 schema, ask "deploy scaffold?"]

STEP 3 — STAFF-NOTE → CUSTOMER-UPDATE (10 min)
- Claude prompt: takes a terse internal note ("primer dry, install Tue PM") + the job context → returns a customer-friendly update ("Materials are prepped and your crew is scheduled for Tuesday afternoon — we'll text 30 min before arrival").
- Auto-detects customer language from the customer name + previous interactions; outputs in matching language.
[GATE 3 — show 3 sample staff-notes + their customer-friendly translations, ask "voice right?"]

STEP 4 — INGEST + RENDER (10 min)
- Run ingestion: insert the 5 sample jobs with their stages + staff notes.
- For each, generate the customer-facing version. Verify rendering at /track/:id for each.
[GATE 4 — show 5 tracking URLs, ask "proceed to deploy?"]

STEP 5 — DEPLOY + DEMO (10 min)
- Deploy via OAuth wrangler.
- Tell human: "Open one of the 5 tracking URLs on your phone. Show: search by tracking ID, see status timeline, see latest customer-friendly update, see photos. Switch to admin view, post a new staff note, watch it translate."
- Loom this 60-second flow.
[GATE 5 — confirm Loom recorded, ask "ready to draft pitch?"]

STEP 6 — DRAFT THE COLD PITCH (5 min)
- Email subject: "How often does someone call asking where their [sign / cabinet / repair] is in the queue?"
- Email body: 4 sentences, lead with the call-volume waste math, attach Loom, link demo URLs, name $5k 7-day pilot / $900/mo, one yes/no close.
[GATE 6 — show draft, ask the human to send manually]

STEP 7 — SHIP THE SUMMARY
- Single-line: "[BUSINESS] status portal live · 5 sample jobs · demo: [LOOM] · pitch sent."
- Append to /home/eratner/cafecito-ai/portal-shipped.md.

DONE.

GUARDRAILS: NEVER use real customer PII in the demo. The 5 sample jobs must be either anonymized or synthetic. Staff-note translation must pass an "owner-eye" check before any go to a customer. Cost ceiling: $3.