Prompt technique extracts writing personas from Slack history to replicate user voice across channels
A prompt tip circulating on X instructs AI agents to mine past messages and build per-recipient communication skills.
Developer @jxnlco shared prompt tip 13 on X, demonstrating how to instruct an AI agent to read a user's past 400 Slack messages, identify distinct communication personas for different contacts, and encode those patterns into reusable skills. The same technique then extends to email and X, giving the agent a channel-aware model of how the user writes.
The approach surfaces a practical pattern in agentic workflows: rather than giving an AI a static style guide, it extracts behavioral signals from real historical data to build per-recipient communication profiles. For builders, this points toward AI agents that adapt tone automatically based on relationship context — not just a single "write in my voice" instruction. For buyers and users, it raises the question of what access an agent needs to be genuinely useful as a communication proxy, and how much historical data is enough to capture voice reliably.