subagents-orchestration-guide

The Subagents Orchestration Guide acts as a central conductor for AI-assisted software development, enabling the seamless coordination of specialized agents. This skill is designed for developers who need to manage complex, multi-layered features that require more than simple code generation. By implementing a strict separation between orchestrator logic and specialist implementation, it ensures that high-level planning, requirements analysis, technical design, and automated testing are handled by dedicated agents, leading to consistent and reviewable outputs.

• Coordinates a library of 15+ specialized agents including requirement-analyzer, prd-creator, task-executor, and quality-fixer to handle the full development lifecycle.

• Enforces structured workflows for small, medium, and large-scale tasks, ensuring that documentation (PRDs, ADRs, Design Docs) is generated before implementation begins.

• Implements autonomous requirement change detection, automatically restarting workflows when scope-expanding signals like new features or constraints are identified.

• Provides a robust delegation boundary model that prevents agents from prescribing implementation details, allowing individual specialists to make autonomous decisions based on project-specific repo conventions and configurations.

• Integrates cross-layer consistency checks using design-sync for fullstack monorepo projects, ensuring that backend and frontend designs remain synchronized.

• Designed for use within Claude Code environments to manage complex feature implementation, refactoring, and reverse engineering tasks.

• Expects input in the form of high-level user requirements or features; outputs a structured pipeline of design documents, work plans, and verified code.

• Employs explicit stop points for user verification, ensuring that the human developer maintains control over product requirements, UI specifications, and architectural design decisions before final implementation.

• Operates best with existing project structures where repository conventions are established, allowing specialists to dynamically determine execution methods (e.g., testing frameworks, linting rules, or build commands) through analysis of the codebase and configuration files.