email-draft

Email Draft Generator

Purpose

Intelligent email draft generator that creates high-quality, style-matched draft replies using a 3-stage workflow. Combines email search, context analysis, and executive writing style profiles to generate draft replies that match the recipient's communication preferences.

Key Features:

• Email Search: Natural language search with fallback strategies
• Context Analysis: Extracts key points and themes from original email
• Style Matching: 6-level fallback algorithm with 4 executive profiles
• Draft Generation: Claude-generated drafts matching recipient style
• MCP Integration: Uses agent365-mail-proxy for email operations

MCP Servers Used

• agent365-mail-proxy - Email search, retrieval, draft creation (included in plugin package)

Mandatory Workflow: 3-Stage Process

Stage 1: Email Search & Discovery

Purpose: Locate the email requiring a reply using natural language search.

Search Strategies (3 strategies with fallback):

1. Direct Email ID (Highest confidence: 100%)

   • User provides exact message ID
   • Fastest, no ambiguity
   • Example: email_id: "AAMkAGI2T..."

2. Natural Language Search (High confidence: 85-95%)

   • User provides search query with sender/subject/keywords
   • Uses Agent 365 natural language processing
   • Example: query: "email from Charles about Q4 budget"

3. Most Recent Email (Fallback: 50-70%)

   • Returns most recent email from inbox
   • Used when search returns no results
   • Example: query: "latest email"

MCP Tool Integration:

# Strategy 1: Direct Email ID
if email_id:
    message = mcp__agent365_mail_proxy__get_message(id=email_id)

# Strategy 2: Natural Language Search
else:
    results = mcp__agent365_mail_proxy__search_messages(
        query=user_query,  # e.g., "emails from Sarah about budget"
        top=5
    )
    # Parse results and select most relevant match
    message = results[0] if results else None

# Strategy 3: Fallback to latest
if not message:
    results = mcp__agent365_mail_proxy__search_messages(
        query="recent emails",
        top=1
    )
    message = results[0] if results else None