With AI agents increasingly acting as procurement/comparison tools, I've been working on this problem and built a prototype: a structured file format (similar to llms.txt but for service capabilities) plus YAML manifests with evaluation data. A demo agent shows how it works: demo.openserviceprotocol.org
But I'm not sure this is the right approach. Has anyone solved this differently? Are there existing standards I'm missing? Particularly interested in hearing from people who run service businesses or build agent frameworks. Has anyone made their business services "agent-readable"?