What problem does it solve? Designing or modifying multi-agent team packages often leads to duplicated rules, conflicting ownership across agent.md, team.md, and config files, and unvalidated changes. This Skill provides a disciplined create/update workflow that keeps every rule in one canonical owner and produces a durable, validated result artifact. ## Core Features & Use Cases - Two operation modes: Create new agent-team packages or update existing ones with a recorded update intent such as optimize, repair, or extend. - Ownership contract enforcement: Assigns each concern to exactly one canonical file (agent.md, agent-config.json, SKILL.md, team.md, team-config.json, references, templates, README) to prevent conflicting definitions. - Validation and recovery: Parses changed JSON, checks frontmatter and skill name alignment, resolves links and handoff addresses, and classifies failures as in-scope corrections, requirement gaps, design impacts, or blockers. - Use Case: When asked to add a new specialist agent to an existing team, the Skill reads the full package topology, applies the smallest coherent delta to the canonical owners, reconciles team-config.json routing, validates the diff, and hands off a persisted result artifact. ## Quick Start Use the agent-team-architecture skill to create a new agent team package for code review with a coordinator and two specialist agents.
