send-rate-update - SKILL.md Agent Skill

name: send-rate-update description: "Paste mortgage rates into chat and send a rate update — publishes a rate page to the website and sends teaser emails via Mailchimp."

You are sending a weekly rate update for Adam Styer's mortgage website. The user pastes rates into chat and you run the full pipeline: normalize rates, preview AI-generated content, confirm, then publish and send emails.

Step 1: Parse Rates

Normalize whatever format the user pastes into:

Product Name: Rate | APR: AprValue

Example:

30-Year Fixed: 6.875% | APR: 6.95%
15-Year Fixed: 6.25% | APR: 6.35%
FHA 30-Year: 6.5% | APR: 7.15%
VA 30-Year: 6.25% | APR: 6.45%

If APR is not provided, omit the | APR: portion.

Step 2: Extract Optional Context

direction: down, up, flat, volatile — infer from user's message
blurb: any market commentary or talking points the user mentions
notes: any additional instructions
audiences: default ["borrower", "realtor"] — narrow if user says "just borrowers" or "realtors only"

Step 3: Confirm Before Sending

Show the user:

Here's what I'll send:

Rates:
[formatted rates]

Direction: [value or "not specified"]
Audiences: [Borrower, Realtor]
Talking points: [blurb or "none"]
Notes: [notes or "none"]

Running preview first — ready?

Wait for confirmation.

Step 4: Preview Call

Use bash tool:

curl -s -X POST https://styermortgage.com/.netlify/functions/generate-rate-update \
  -H "Content-Type: application/json" \
  -d '{
    "rates": "<rates string>",
    "direction": "<direction>",
    "blurb": "<blurb>",
    "notes": "<notes>",
    "audiences": ["borrower", "realtor"],
    "mode": "preview"
  }'

Step 5: Show Preview Results

From JSON response, display:

preview.pageTitle — page title
preview.webContent — summarize HTML as readable text
preview.borrowerSubject — borrower email subject line
preview.realtorSubject — realtor email subject line
pageUrl — where it will publish

Ask: "Ready to go live? This publishes the rate page and sends emails."

Step 6: Live Call

Same payload, "mode": "live":

curl -s -X POST https://styermortgage.com/.netlify/functions/generate-rate-update \
  -H "Content-Type: application/json" \
  -d '{
    "rates": "<same rates>",
    "direction": "<same direction>",
    "blurb": "<same blurb>",
    "notes": "<same notes>",
    "audiences": ["borrower", "realtor"],
    "mode": "live"
  }'

Step 7: Confirm Results

Report:

Page published at pageUrl
Each campaign from campaigns array — audience + status
Flag any errors

Step 8: Log to Marketing History

After a successful live publish, log the rate update to the LoanOS History tab so cadence badges stay current.

Read the secret from .env.local:

AGENT_SECRET=$(grep LOANOS_AGENT_SECRET /Users/adamstyer/Documents/loanos-clone/.env.local | cut -d'"' -f2)

Then log:

curl -s -X POST https://loanos-self.vercel.app/api/marketing/log \
  -H "Content-Type: application/json" \
  -H "X-Webhook-Secret: $AGENT_SECRET" \
  -d '{
    "activity": "Rate update sent — <date>",
    "channel": "Rate Update",
    "notes": "<first 120 chars of rates string>",
    "tracker": "rate-update"
  }'

Replace <date> with today's formatted date (e.g. "Apr 3, 2026") and <notes> with a summary of the rates sent. The tracker field updates the cadence badge on the History tab.

If the curl fails, report the error but don't block — the rate update itself already went out.

Error Handling

API error → show message, offer retry
Empty/unparseable rates → ask user to re-paste
Network failure → suggest checking if Netlify deployment is complete

Notes

Always preview before live — never skip.
Rate page URL pattern: https://styermortgage.com/rates/YYYY-MM-DD.html
No auth required — keys are Netlify environment variables server-side
Emails sent via Mailchimp to borrower and/or realtor lists