Agent Team Orchestration

**Judgment:** A strong, production-minded orchestration skill for multi-agent teams, best suited for sustained workflows rather than one-off delegations. It offers clear roles, a concrete task state machine, structured handoffs, and mandatory review steps, which together materially improve coordination and quality in multi-agent setups. **Key strengths:** - Enforces a simple but powerful pattern (orchestrator + builder + reviewer) that scales into more complex teams. - Explicit lifecycle and comments at transitions make it much easier to debug stuck or failed work. - Standardized handoff format and shared artifact paths directly address the most common multi-agent failure modes (lost artifacts, vague “done” states, silent agents). **Risks / limitations:** - Adds non-trivial process overhead; overkill for single-agent or very small, one-off tasks. - Still relies on correct implementation of external task storage and artifact paths; misconfiguration there will undermine many of its benefits. - Does not itself provide advanced scheduling, load balancing, or strict SLAs—those must be layered on top if needed. **Recommended scenarios:** - Long-lived, multi-step projects where multiple agents depend on each other’s outputs (e.g., ongoing product development, complex research → implementation loops). - Teams that have already hit coordination problems (lost work, unreviewed changes, unclear ownership) and need more disciplined orchestration. - Any environment where quality and traceability matter more than raw throughput, and where an orchestrator agent (or human) can own task state transitions and oversight.