agent-selector

Agent selection automates mapping your PRD's technical architecture—languages, frameworks, databases, and infrastructure—to specialized agents. Trigger the Skill after Phase 5 task decomposition to generate agent-config.json and populate your .claude/agents/ directory with matched configurations.

Agent selection reads PRD Section 8 Technical Architecture to extract your tech stack details. It analyzes languages, frameworks, databases, infrastructure components, plus security and scalability requirements to map each to appropriate agents.

Yes. Agent selection evaluates security and scalability demands from your PRD and selects auditor and scale-focused agents accordingly. The generated config reflects both functional requirements and non-functional constraints.

Agent selection generates two outputs: .claude/agent-config.json with local and upstream agent strategy, and a populated .claude/agents/ directory with agent configurations ready for Phase 6 specification generation.

Run agent selection after Phase 5 task decomposition completes and before Phase 6 specification generation starts. Activate with [ACTIVATE:AGENT_SELECTOR_V1] to enforce phase-gate sequencing and deployment readiness.

Agent selection works with tech stacks defined in your PRD's technical architecture section. It maps common languages, frameworks, databases, and infrastructure types to agents; the explicit tech-stack mapping in your repository determines supported combinations.