navotvolkgroundup/.claude/skills/x-reply

X Reply Generator for @nabotweet

Overview

Generate replies to people who respond to your tweets. Replies should sound natural, match your voice, and add value to the conversation.

Before Writing

1. Read context profiles in /context/:

   • voice.json — How to sound
   • audience.json — Who you're talking to
   • business-groundup.json — GroundUp Ventures context
   • business-weeklysync.json — Weekly Sync podcast context

2. Read replies data from /knowledge/engagement/replies.json

   • If no replies file exists, run: python3 scrape_replies.py

3. Check engagement data in /knowledge/engagement/ for context on what's working

Reply Pipeline

Step 1: Scrape replies

Run python3 scrape_replies.py to get fresh replies.

Step 2: Analyze each reply

For each reply, understand:

• What are they saying? Agreement, question, pushback, joke?
• Do they have a point worth engaging with?
• Is this someone worth building a relationship with?
• Do I need to research something to reply well?

Step 3: Research if needed

If a reply references something specific (a company, a trend, a claim), research it before replying. Use firecrawl to look it up.

Step 4: Write replies

Generate a reply for each meaningful response. Skip:

• Generic "great post" / emoji-only replies
• Spam / self-promotion
• Trolls not worth engaging

Step 5: Humanize (MANDATORY)

Run every reply through the humanizer skill before presenting. Check for AI patterns.

Step 6: Queue for approval

Add each reply using: python3 bot.py reply-add "reply text" tweet_url

Reply Types

Acknowledgment (for agreement/praise)

• Short, warm, no ego
• Add a small insight or joke if possible
• Don't just say "thanks"

Answer (for questions)

• Direct answer first
• Add context if useful
• If you don't know, say so

Pushback response (for disagreement)

• Acknowledge their point
• Share your perspective without being defensive
• "Fair point, but..." is fine. "You're wrong" is not.

Banter (for jokes/humor)

• Match their energy
• Hebrew humor when appropriate
• Keep it light

Voice Rules (same as x-tweet)

ALWAYS

• Sound like someone in the room, not behind a podium
• Natural Hebrew-English code-switching
• Be generous with knowledge
• Short replies (1-2 sentences ideal)
• Match the language of the person replying (Hebrew reply = Hebrew response)

NEVER

• Corporate PR tone
• Defensive or dismissive
• "Great question!" or other sycophantic openers
• Long-winded explanations when a sentence will do
• Ignore the person's actual point
• Generic motivational responses

Language

• Reply in the same language as the person
• If they write Hebrew, reply Hebrew
• If they write English, reply English
• Hebrew-English mix follows natural patterns

Length

• Ideal: 1-2 sentences
• Max: 3 sentences for complex answers
• Shorter is almost always better

Quality Checklist

Before delivering:

• [ ] Does this reply add value to the conversation?
• [ ] Does it sound like a person, not a brand?
• [ ] Is it in the right language?
• [ ] Has it been humanized?
• [ ] Would I actually say this in person?