What problem does it solve?
Multi-agent platforms struggle with brittle discovery, identity/location conflation, undocumented lifecycle transitions, and undifferentiated failures that cause race conditions, stale routing, and failed recoveries. This Skill codifies the FIPA Agent Management patterns to separate identity from address, split authoritative lifecycle/identity management (AMS) from non-authoritative capability directories (DF), and provide a structured exception and lifecycle model so orchestrators can discover, invoke, and recover agents reliably.
Core Features & Use Cases
- AMS / DF separation: Prescribes a white-pages AMS for authoritative lifecycle and identity and a yellow-pages DF for capability advertisement and federated search.
- AID-based resolution & federated search: Use immutable AIDs resolved via resolver chains and depth-bounded DF federation for scalable discovery across organizational boundaries.
- Lifecycle & exception contracts: Formalizes six lifecycle states with transition ownership and a rich exception taxonomy (refuse/failure/not-understood) to enable deterministic retry and recovery strategies.
- Use case: An orchestrator queries the local DF for a translation capability, falls back to federated DFs when needed, resolves the chosen AID via the AMS to obtain transport addresses, checks lifecycle state before invocation, and applies exception-specific recovery (re-discover, escalate, or treat as idempotent success).
Quick Start
Find active agents that advertise the "translation" service across federated directories, resolve their AIDs to current transport addresses, and return the preferred reachable endpoint for each agent.