What problem does it solve? Getting a well-grounded answer to a complex question often requires multiple perspectives and current web evidence, but a single agent response can be biased, outdated, or unverified. This Skill orchestrates a fixed three-agent General Council (generalist, skeptic, domain expert) that researches the web first, deliberates in parallel, reconciles disagreements, and produces a synthesized advisory answer. ## Core Features & Use Cases - Curated Web Research Phase: Runs targeted web_search queries with freshness and temporal-intent metadata before dispatching any agents, and blocks dispatch when time-sensitive questions lack current sources. - Parallel Multi-Agent Deliberation: Dispatches generalist, skeptic, and domain expert lanes asynchronously, collects results, and routes disagreements back for one targeted reconciliation round. - Structured Synthesis Output: Leads with consensus, honestly presents persisting disagreements, cites the strongest sources, and marks unverifiable claims with low confidence. - Use Case: A user runs /swarm council --spec-review while drafting a technical spec; the council researches current library behavior, three agents independently assess the design, disagreements are reconciled, and the user receives a cited advisory summary without any code or spec being modified. ## Quick Start Ask the architect to run /swarm council followed by your question, for example to get an advisory multi-perspective answer on a design decision.
