Create engineering-email-intelligence-engineer.md

[Sammy-spk]

PR Description: Email Intelligence Engineer Agent

Agent Information

Agent Name: Email Intelligence Engineer
Category: engineering
Specialty: Extracting structured, reasoning-ready data from raw email threads for AI agents and automation systems

Motivation

None of the current engineering agents address the specific domain of turning unstructured email data into structured context for AI agents. The AI Engineer covers ML broadly, and the Backend Architect handles general API/infrastructure design, but email intelligence has its own unique failure modes and engineering challenges.

Anyone who's tried to build an AI agent that works with email data knows the gap: Gmail API gives you raw MIME, but your agent needs structured understanding (who decided what, what was committed, which attachments matter). The pipeline between those two points, thread reconstruction, quoted text deduplication, participant role detection, attachment extraction, hybrid retrieval, context assembly is substantial engineering work that's distinct from general ML or backend architecture.

This agent fills that gap with practical patterns for the full pipeline: from raw email ingestion through to structured tool interfaces that agent frameworks can consume.

Disclosure

I work on the growth team at iGPT which builds a Context Intelligence API for email data. The agent references iGPT as one option alongside building your own pipeline and using other tools. The agent is designed to be genuinely useful regardless of what tools someone chooses, the email processing patterns, failure modes, and architecture guidance apply universally.

Testing

• Patterns in this agent are drawn from real-world email processing challenges: thread reconstruction across providers, quoted text deduplication ratios, participant detection accuracy benchmarks
• Code examples are functional Python patterns (not pseudocode) that demonstrate the actual engineering involved
• Failure modes referenced (forwarded chain collapse, CC drift, attachment orphaning, quoted reply duplication) are well-documented problems in email processing

Checklist

• Follows agent template structure (frontmatter table, identity, mission, rules, capabilities, workflow, communication style, success metrics, advanced capabilities)
• Includes personality and voice
• Has concrete code/template examples
• Defines success metrics with quantitative benchmarks
• Includes step-by-step workflow
• Proofread and formatted correctly
• Tested in real scenarios

[msitarzewski]

Hey @Sammy-spk — the domain here is solid. Email intelligence for AI agents is a real gap and the content (MIME parsing, thread reconstruction, hybrid retrieval, failure modes) is genuinely useful. Two things to fix before we can merge:

1. Frontmatter format

The file uses a markdown table for frontmatter — we need YAML with emoji and vibe fields. Should look like:

---
name: Email Intelligence Engineer
description: Expert in extracting structured, reasoning-ready data from raw email threads for AI agents and automation systems
color: indigo
emoji: 📧
vibe: Turns messy MIME into reasoning-ready context — because raw email is noise and your agent deserves signal
---

(Pick whatever emoji/vibe feels right — just needs to be there.)

2. Vendor-specific code example

We appreciate the upfront disclosure about iGPT — that's the right way to handle it. But the Step 4 code sample with from igptai import IGPT reads as a product demo embedded in the agent. Our editorial bar is that agents should stand alone as prompts without promoting a specific vendor's SDK.

Could you replace the iGPT-specific example with a generic pattern? The "Option A: Build it yourself" and "Option C: Use framework-specific integrations" sections are great — you could expand one of those instead. Mentioning that context intelligence APIs exist as a category is totally fine, just without naming a specific product or showing its import.

Everything else looks good — looking forward to the update!

[Sammy-spk]

Thanks for the feedback. Updated the frontmatter to YAML with emoji/vibe and replaced the vendor-specific example with a generic implementation pattern so the agent stands alone as a reusable prompt. Let me know if anything else should be adjusted.