Linkedin Message Writer

linkedin-message-writer

Purpose: Generate LinkedIn messages that sound like a real human wrote them - direct opener, specific reason for reaching out, one clear ask.

Input Schema

{
  "message_type": "outreach",
  "recipient_name": "Priya",
  "context": "Priya is an eng manager at Stripe, posted about their platform team, I'm interested in staff eng roles",
  "ask": "15-minute call to learn about how the platform team is structured",
  "your_name": "Nick",
  "tone": "direct"
}

Output Schema

{
  "message": "...",
  "subject": "optional - for InMail",
  "word_count": 65,
  "patterns_removed": [7, 19, 21]
}

Prompt Flow

Pass 1: One specific hook, one clear ask, sign-off. Under 75 words. Remove: chatbot openers (P19), sycophancy (P21), promotional adjectives (P4), AI vocab (P7), filler (P22).

Pass 2: Does it pass the "would I cringe if I received this?" test? Remove any remaining flattery or AI tells.

Examples

Short - cold outreach

Before: "I hope this message finds you well! I am incredibly excited to connect with you and would love to potentially explore some opportunities." After: "Hi Priya - saw your post on Stripe's platform team. I'm a staff eng focused on infra and developer experience. Would a 15-minute call make sense to talk through how you've structured the team? Happy to share what I've been working on in return."

Medium - response to a job post

Before: "Great opportunity! I am passionate about leveraging my extensive experience to add value to your dynamic team and contribute to your transformative mission." After: "Hi - I saw the staff eng posting. I've spent 3 years building out platform tooling (CI, feature flags, deploy automation) at Series B and Series C companies. Happy to send a resume if it looks like a fit."

Long - referral ask

Unit Tests

# tests/skills/test_linkedin_message_writer.py
from ai_pattern_scrubber import detect_patterns

OUTREACH = "Hi Priya - saw your post on Stripe's platform team. I'm a staff eng focused on infra and developer tooling. Would a quick call make sense? Happy to share what I've been working on."
RESPONSE = "Hi - I saw the staff eng posting. 3 years of platform tooling work (CI, feature flags, deploy automation). Happy to send a resume if it's a fit."

def test_outreach_no_sycophancy():
    assert not [h for h in detect_patterns(OUTREACH) if h.id == 21]

def test_response_no_promotional_language():
    assert not [h for h in detect_patterns(RESPONSE) if h.id == 4]

def test_outreach_no_chatbot_artifacts():
    assert not [h for h in detect_patterns(OUTREACH) if h.id == 19]