Harmony: Media-Carried Workflow Signals for Distributed AI Agent Execution
Year: 2026
Type: preprint
Abstract: Harmony embeds machine-readable workflow signals in the audio track of media and connects playback to approved agent work. The embedded cue carries a routing identity (organization, workflow, channel, agent suite) rather than a fixed task, so a media asset authored once can address a task queue whose contents evolve over time. This paper describes the complete prototype system, its signal layer, its security and privacy posture including known weaknesses, and a five-part validation program. Preprint, not yet peer reviewed. The system described is patent pending.
Author: Cheeks, Milan
Institution: University of Colorado Boulder
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Keywords: task queues, privacy and security
Open Access status: green
OpenAlex ID: W7236781722
