Adding a warm intro request generator to the job application kit

The job application kit now has a fifth output: a warm intro request.

Why referrals matter

Employee referrals convert 3–5x better than cold applications. The mechanics are simple: referred candidates skip the résumé screening stage, because someone the hiring team trusts has already pre-screened them. The problem is that most people don't know how to ask.

A bad referral ask: "Hey, I saw your company is hiring — can you refer me?" It puts the connection in an awkward position (they don't know what role, they don't know if you're a fit), and it gives them nothing to work with if they forward it internally.

A good referral ask is specific, short, and easy to say yes to. It mentions the exact role. It gives one reason you're a fit. It includes an explicit "no pressure" exit ramp so the connection doesn't feel obligated if it's not a fit.

What the endpoint does

POST /api/warm-intro takes:

• Job description (required)
• Connection name (required)
• How you know them (optional: "Former colleague at Startup X")
• LinkedIn URL (optional)

It returns:

{
  "subject": "Quick question about the Backend Engineer role at Acme",
  "message": "Hey Sarah — quick question if you have a moment.\n\nI saw Acme is hiring a Senior Backend Engineer and it looks like a strong match — I've been building Python serverless systems on AWS for the last three years...",
  "follow_up": "Hi Sarah — just following up on my message from last week...",
  "connection_value": "Strong — Sarah currently works at Acme and can submit a direct employee referral."
}

The connection_value field is the one I find most useful: it tells you upfront whether this connection is well-positioned to help (currently works there, has an active referral system) versus weakly positioned (left the company two years ago, might have a warm introduction but not referral access).

How Haiku generates the message

The prompt gives Haiku the engineer's full CV, the job description, and the connection info, then instructs it to write within specific constraints:

• Under 150 words (longer messages get skimmed)
• No "I hope this message finds you well" opening
• One specific detail from the CV as the hook
• An explicit "no pressure" sentence
• A follow-up that doesn't re-pitch, just nudges

I added the follow-up generation because the most common mistake after sending a referral ask is either following up too aggressively or not at all. Having the follow-up drafted upfront removes the decision friction.

Where it fits in the dashboard

The warm intro section lives below the cover letter generator in the Pro dashboard. You paste the JD, enter the connection's name and how you know them, and click "Draft warm intro →". The UI shows the connection value assessment first (so you know before reading the message whether this contact is worth asking), then the message and follow-up in copyable boxes.

This is the fifth Pro AI endpoint:

1. POST /api/tailor — tailor CV sections to a specific JD
2. POST /api/gap-analysis — identify skill gaps for a specific role
3. POST /api/interview-prep — generate 5 grounded Q&A pairs
4. POST /api/cover-letter — write a tailored cover letter
5. POST /api/warm-intro — draft the referral ask

The job application kit runs all four of the first endpoints in parallel from a single JD paste. The warm intro is separate because it needs the additional connection context.