spawn-decision

The spawn-decision skill integrates the decision-agent into the multi-agent orchestration workflow. It serves as a high-level reasoning layer triggered when the primary orchestrator requires expert judgment to maintain efficiency and architectural integrity. By utilizing the Opus model, it evaluates complex project states, determines the necessity of Plan Mode, and resolves recurring technical obstacles that standard agents cannot overcome.

• Performs deep-dive architectural analysis and compares implementation strategies for technical stacks, design patterns, and framework integration.

• Evaluates task complexity to decide between immediate execution or transitioning to a formal Plan Mode for complex, multi-step requirements.

• Provides automated root-cause analysis for recurring errors, leveraging existing LESSONS_LEARNED patterns to suggest actionable resolutions.

• Acts as a smart gateway for user input, transforming ambiguous or high-stakes requests into clear, actionable options for project direction.

• Streamlines cost optimization by ensuring expensive model calls are reserved only for scenarios that demand high-level cognitive reasoning.

• Input consists of decision types (plan_mode, architecture, error_resolution), situational context, search engine summaries, and current technical dependencies.

• Expected output is a structured recommendation (e.g., ENTER_PLAN_MODE, EXECUTE, NEED_MORE_INFO) supported by concise, evidence-based reasoning.

• Operates strictly within the .claude/agents/ orchestration framework, ensuring consistency with project-specific settings and conventions.

• Designed to reduce hallucination and unnecessary context bloating by focusing on domain-specific logic and constrained problem-solving paths.

• Requires careful management of context to ensure the decision-agent has sufficient information (codebase search results, error logs) to make high-confidence decisions.