bmad-agent-builder

You can create an AI agent by describing the agent you want to build, which triggers conversational discovery to define its identity, capabilities, memory, and activation routing. The process outputs a ready-to-use Skill directory with SKILL.md, references, assets, and scripts.

Stateless, memory, and autonomous agent archetypes differ by their distinct sanctum and activation flows. The builder classifies your agent into one of these three types during guided discovery based on whether it requires persistent memory or autonomous decision-making.

Yes, you can analyze an agent or edit an existing agent by requesting those actions directly. The builder orchestrates the end-to-end creation, editing, and analysis of Agent Skills through conversational discovery to update or inspect the skill structure.

Building an agent skill generates a complete directory structure including SKILL.md, references, assets, and optional scripts for deterministic tasks. These components provide the ready-to-use foundation for the agent's persona and activation routing.

No, you do not need to define memory and activation routing beforehand. The guided conversational discovery process prompts you to establish the agent's identity, capabilities, and memory requirements dynamically during the build session.

You should use scripts in an autonomous agent skill when you need to execute deterministic tasks. The builder includes optional scripts in the output directory alongside SKILL.md and references to handle these specific automated functions.