bmad-agent-builder

Guided discovery for AI agent design walks teams through conversational prompts to identify agent identity, capabilities, memory needs, and activation flow. This process outputs a complete agent skill structure with identity seeds, sanctum templates, and capability prompts for domain-specific workflows.

Conversational discovery supports converting existing agents, editing behaviors, and rebuilding from intent. It produces a lean bootloader with sanctum templates, using progressive disclosure to finalize the complete agent skill structure for memory-enabled assistants.

Building memory-enabled assistants and coaching copilots requires identifying memory needs and activation flow through guided discovery. The process outputs configuration for both headless and memory modes, creating sanctum templates and capability prompts tailored for education, software, and business workflows.

The agent builder requires the pyyaml Python dependency to function. It uses scripts, references, and assets to guide teams through creating domain-specific agents, memory-enabled assistants, and coaching copilots across various workflow contexts.

Progressive disclosure refines AI agent capabilities by gradually revealing complexity during conversational discovery. It starts with a lean bootloader and sanctum templates, then progressively finalizes the complete agent skill structure with identity seeds and capability prompts.

Sanctum templates and identity seeds provide the foundational architecture for AI agents by establishing core identity and memory structure. They enable progressive disclosure to finalize capability prompts and configuration for headless and memory modes across domain-specific workflows.