Resolver
person concept tool org talk claim — click a node to jump to its page; hover an arrow for the relation
A Resolver is a routing/dispatch table used in multi-agent systems that maps available agent capabilities to their descriptions, functioning similarly to an agents.md file listing what each agent can do. It serves as the lookup mechanism by which an orchestrating system determines which skill or agent to invoke for a given task.
Role in Agent Architecture
The Resolver acts as the central registry of agent capabilities. As described in the YC AI Playbook talk, it works by "plugging into the resolver which is like your agents.md with like the list of things that the agents can do" — Inside YC's AI Playbook, 15:51. Skillify is one system described as using this pattern, feeding its skill definitions into the Resolver so the orchestrator can select appropriate capabilities at runtime.
Design Principles: DRY and MECE
The quality of a Resolver depends heavily on how cleanly its entries are defined. Two principles are emphasized as critical:
- DRY (Don't Repeat Yourself): When adding new skills, the system should "look at all of the other skills and tools that exist and is it you know dry don't don't repeat yourself" — Inside YC's AI Playbook, 16:41. Overlapping or redundant entries degrade routing quality.
- MECE (Mutually Exclusive, Collectively Exhaustive): Resolver entries should cover the full capability space without overlap. The claim is that "if you have a dry and MECE resolver table anywhere it's actually like the optimal resolver" — Inside YC's AI Playbook, 16:44.
In practice, this means preferring one parameterized skill over ten narrowly overlapping skills, keeping the Resolver table compact and unambiguous for the routing model to navigate.