genudo-pre-ai-channel-audit

GenuDo Pre-AI Channel Discovery & Arabic Executive PDF Audit

Overview

Use this skill whenever a client connects new social or messaging channels (Facebook Messenger, Instagram DM, WhatsApp) to GenuDo Pipelines in passive/observation mode before launching the AI Agent.

This workflow logs into GenuDo via MCP, extracts all real visitor messages and leads from the newly connected channels, analyzes what customers are asking about, measures each inquiry category's share of total volume, projects 30-day message/lead volume and expected monthly cost, defines the AI Agent's use cases, and compiles an Executive Arabic PDF Report (RTL, 10 pages, one case per page, locked A4 layout with zero text overflow).

---

1. Prerequisites & GenuDo MCP Login / Discovery

Before running the analysis, authenticate and verify access to the target GenuDo workspace via the GenuDo MCP Server (https://api.genudo.ai/mcp).

Step 1.1: Verify GenuDo Authentication

1. Call get_profile and get_account_summary from the genudo MCP server to confirm your active session and account permissions.
   • If the MCP server is not yet connected or returns 401 Unauthorized, configure the genudo MCP server in your client settings (mcp_config.json or .claude.json) with your GenuDo API token / OAuth login, then re-verify with get_profile.
2. Call list_pipelines to inspect all pipelines in the workspace.

Step 1.2: Identify Newly Connected Channels & Pipelines

For each target pipeline (e.g., Online Store vs. Publishing House, or Branch A vs. Branch B):

1. Call get_pipeline_details(pipeline_id=<ID>) and inspect the channels array:

   {
     "id": 301,
     "identifier": "104352818912345",
     "status": "active",
     "provider": "messenger",
     "connector": "facebook"
   }
   

2. Call list_pipeline_stages(pipeline_id=<ID>) to check stage names, order, and whether ai_paused is true (passive data collection mode prior to AI launch).

Step 1.3: Extract All Visitor Messages & Leads (Read-Only)

1. Call list_messages(pipeline_id=<ID>, limit=100) and paginate using next_cursor until all messages in the observation window (typically 7 to 30 days) are fetched.
   • Rate Limit: Respect GenuDo's 30 requests/minute limit. Cache raw payloads locally to messages_raw.json.
   • Read-Only Rule: Never mutate pipelines, stages, or opportunities (update_*, create_*, delete_*) during an audit.
2. Filter out internal test contacts (is_test == true or contact names containing test, autotest, تجربة).
3. Group messages by contact_id (each unique contact_id = 1 Unique Lead / Conversation).
4. Record message metadata:
   • Total messages (COUNT(messages)), Unique leads (COUNT(DISTINCT contact_id)), and Average messages per lead (Total Messages / Unique Leads).
   • Daily breakdown by Cairo local date (UTC+3) to identify average daily flow and peak traffic days.
   • Attachment breakdown (attachments array): count how many messages contain images (e.g., product/book cover photos, payment screenshots), audio/voice notes, PDFs/CVs, or external links.