agency-email-intelligence-engineer

What problem does it solve? Raw email threads are structurally chaotic: quoted replies duplicate content 4-5x, forwarded chains collapse multiple conversations, and flattened threads strip From: headers so action items get attributed to the wrong people. This Skill builds pipelines that convert raw email into clean, structured context that AI agents can reason over reliably. ## Core Features & Use Cases - Thread Reconstruction: Resolve In-Reply-To/References header chains to rebuild conversation topology across replies, forwards, and forks. - Quoted Content Deduplication: Strip prefix-based, delimiter-based, and Outlook XML quoting to reduce token bloat by 4-5x without information loss. - Participant & Action Item Extraction: Bind commitments and decisions to the correct sender using preserved message-level identity. - Hybrid Retrieval & Context Assembly: Combine semantic search, full-text search, and metadata filters with token-budget-aware context assembly and source citations. - Use Case: Connect a Gmail or Microsoft Graph mailbox, index historical threads, and expose LangChain tools like email_ask and email_search so an agent can answer questions with citations grounded in specific messages. ## Quick Start Ask the agent to reconstruct and deduplicate a raw email thread, then extract the participant map, decisions, and action items as structured JSON with source citations.