orchestrator

The orchestrator skill serves as the central command for Multi-Agent Ralph, an autonomous development framework designed to extend Claude Code's capabilities. It manages a comprehensive 10-step lifecycle—including evaluate, clarify, classify, plan, execute, and retrospective—to handle complex software tasks that exceed the capacity of single-agent interactions. By utilizing Aristotle First Principles for task deconstruction and a MemPalace-inspired memory architecture, the skill ensures that both trivial bug fixes and high-complexity refactoring or multi-file changes are executed with structural rigor and logical consistency.

• Swarm mode capability: Automatically spawns a team of specialized agents, including coder, reviewer, tester, and security-auditor, to execute tasks in parallel.

• Memory-driven lifecycle: Integrates persistent memory stacks (L0-L3) to leverage learned project taxonomy and institutional knowledge across sessions.

• Quality gate enforcement: Implements mandatory validation stages, covering code correctness, security auditing (semgrep/gitleaks), and adherence to performance standards.

• Aristotle-based analysis: Employs first-principles deconstruction to identify high-leverage actions and remove inherited assumptions before task execution.

• Dynamic model routing: Automatically routes tasks to the appropriate model (e.g., Opus, Sonnet, or GLM) based on complexity scores from 1 to 10.

• Invoke the skill using /orchestrator followed by the task description to trigger the automated workflow.

• Use the --teammates flag to explicitly define the agent team composition for custom coordination needs.

• Designed for complex engineering workflows such as migrating database schemas, implementing major authentication features, or performing large-scale code refactors.

• Inputs typically include high-level feature requests or technical problem statements; outputs comprise a documented plan, parallel sub-agent executions, validation reports, and a final retrospective.

• Practical constraints: Ensure that independent tasks are identified to maximize parallel efficiency; sequential tasks must be explicitly defined via dependency structures to avoid race conditions.