mirror of
https://github.com/hanzoskill/linkedin-inbox.git
synced 2026-10-30 21:12:44 +00:00
3.6 KiB
3.6 KiB
Communication Style Extraction Guide
How to analyze a user's communication patterns and apply them to LinkedIn draft responses.
Step 1: Read Style Source
Check for style information in this order:
USER.md- Usually contains explicit communication ruleslinkedin-inbox-config.json- May have custom style overrides- Historical messages - If available, analyze past sent messages
Step 2: Extract Key Signals
Formality Level
- Formal: "Dear [Name]", full sentences, "Best regards"
- Professional casual: "Hi [Name]", contractions OK, "Thanks"
- Casual: "Hey", fragments OK, emoji allowed
Greeting Patterns
Common patterns to detect:
Hi [Name]/Hey [Name]/Hello [Name]- No greeting (jumps to content)
- First name only vs "Mr./Ms."
Sign-off Patterns
Best/Thanks/Cheers- Name only
- No sign-off
Sentence Structure
- Short, punchy sentences
- Longer, flowing prose
- Bullet points for clarity
- Questions to engage
Vocabulary Preferences
Detect positive patterns:
- Preferred words they use often
- Industry jargon comfort level
- Contraction usage (don't vs do not)
Detect negative patterns (banned):
- Words explicitly called out as banned
- Overly salesy language they avoid
- Corporate buzzwords they skip
Response Length
- One-liner responses
- 2-3 sentence default
- Detailed paragraphs
- Match to context (short for simple, longer for complex)
Step 3: Build Style Profile
Create a mental model like:
Tone: Professional casual
Greeting: "Hi [Name]" or direct start
Sign-off: None or "Thanks"
Length: 2-4 sentences max
Vocabulary: Plain English, no jargon
Banned: "excited", "leverage", "circle back"
Special: Uses colons instead of dashes
Step 4: Apply to Drafts
Template Selection
Based on message type:
- Decline pitch: Polite but firm, 1-2 sentences
- Express interest: Warm but not eager, ask one question
- Defer: Acknowledge, buy time, no commitment
Draft Generation
- Start with appropriate greeting (or none)
- Address their point directly
- State position clearly
- Offer next step (if any)
- Close with user's typical sign-off
Quality Check
Before presenting draft:
- No banned words?
- Matches typical length?
- Sounds like the user wrote it?
- Not too eager or too cold?
- Clear and actionable?
Example Style Profiles
Founder-to-Founder (Low-BS)
Tone: Calm, controlled, non-reactive
Greeting: "Hi [Name]" or skip
Sign-off: None
Length: 3-6 lines max
Banned: excited, game-changer, hop on a call, quick win, leverage
Special: Always leave an exit ("all good if not")
Friendly Networker
Tone: Warm, approachable
Greeting: "Hey [Name]!"
Sign-off: "Cheers" or "Talk soon"
Length: 2-4 sentences
Allowed: Light emoji (😊), exclamations
Special: Ask about them before pitching
Direct Operator
Tone: No-nonsense
Greeting: Skip
Sign-off: Skip
Length: 1-2 sentences
Style: Get to the point immediately
Special: Binary questions ("Yes or no?")
Handling Edge Cases
Unknown Style
If no style info available:
- Default to professional casual
- Keep drafts short (2-3 sentences)
- Avoid anything too bold
- Flag to user: "No style profile found, using defaults"
Conflicting Signals
If USER.md contradicts config:
- Config takes precedence for explicit overrides
- USER.md provides baseline defaults
- Note the conflict in draft: "Using config override for [X]"
Cultural Adaptation
For international contacts:
- Slightly more formal by default
- Full words over contractions
- Clearer structure
- Avoid idioms