Auto Expert Team

Problem: multi-role collaboration is not the default

Many teams treat “AI as product manager, architect, developer” as the default for complex work, but the source frames this as an anti-pattern: AI can handle broad-knowledge, generation, and context-window tasks end-to-end, while roles are mostly presentation and attention bias rather than capability switching. The skill addresses what to do when you still need explicit roles, interfaces, and coordination: how to avoid overlapping responsibilities, weak prompts, and broken collaboration chains.

How it works: task analysis to verification

• Task analysis: identify domains, subtasks, dependencies, and whether teaming is justified.
• Role and collaboration design: define core, collaborating, supporting, and coordinating experts with clear boundaries and interfaces, then generate role prompts and collaboration prompts.
• Execution and verification: follow the R0-01 to R0-06 chain to produce analysis, role, collaboration, assembly, execution, and quality reports.

Boundaries

This is closer to an expert-team framework generator than a replacement for real engineering process. The source emphasizes using the smallest viable structure: one expert before a small team, a small team before a large one. For simple or single-domain tasks, a direct end-to-end prompt is usually better.