Call Transcript Knowledge Base

What it does

Processes call transcripts (coaching, sales, support) and converts them into structured, categorized knowledge notes (frameworks, strategies, common objections, lessons learned). Uses a deep extraction pass with a document analysis model and commits the resulting notes to the knowledge base repo.

When to use it

• Having a backlog of unorganized coaching or sales call history
• Wanting to turn recurring support conversations into an objections-and-responses database
• Preparing training material for a sales team using real calls from GoHighLevel or another telephony tool

Steps

1. Pull the transcripts from the source (Fathom, GoHighLevel recordings, or other)
2. Run the transcript through a deep extraction analysis to identify key themes
3. Classify findings into categories (frameworks, strategies, objections, lessons)
4. Write each note in Obsidian-compatible Markdown
5. Review the generated notes before confirming them
6. Commit the approved notes to the knowledge base repository

The skill

---
name: call-transcript-knowledge-base
description: Process coaching, sales, or support call transcripts into structured Obsidian knowledge notes (frameworks, strategies, objections, lessons).
---

# Call Transcript Knowledge Base

## When to use it
When you have coaching, sales, or support call transcripts and want to turn them into a reusable knowledge base.

## Transcript sources
Recording tools like Fathom, or calls logged through GoHighLevel and other telephony/CRM systems.

## Extraction process
The full transcript is analyzed for recurring patterns: frameworks mentioned, strategies applied, objections raised by the client, and lessons learned.

## Note categorization
Each finding is classified into one of four categories and written as a standalone Markdown note with cross-links where relevant.

## Use in GoHighLevel sales
Documented objections feed call scripts and automated responses within GoHighLevel follow-up workflows.

## Safety rules
Notes are generated as a draft and reviewed before being committed to the repo. Never invent content that isn't in the original transcript. Don't include sensitive client data without anonymizing it if required.

For Claude Code, save it as ~/.claude/skills/call-transcript-knowledge-base/SKILL.md. In Cursor, ChatGPT or another assistant, paste it as project instructions.