bmad-agent-builder

To build an autonomous agent skill, you provide a repository path or existing skill directory. The builder applies BMad standards to generate the correct architecture, capability routing, and SKILL.md metadata for stateless, memory, or autonomous setups.

Yes, you can analyze an existing agent skill for quality, cohesion, prompt craft, and execution efficiency. The builder reviews your agent directory against BMad standards to identify customization issues and validate its design intent.

To edit a BMad agent skill safely, use the targeted revision workflow. It applies improvements to your SKILL.md and supporting files while preserving the existing agent's identity, architecture, and original design intent.

No, you do not need an existing SKILL.md to generate metadata. The builder can create new agent skills from scratch, handling stateless, memory, or autonomous architectures while generating required sanctum scaffolding.

Yes, the BMad skill builder works with standard Python environments. It requires the pyyaml dependency to parse and generate YAML configurations for agent metadata and capability routing across BMad Method repositories.

You should avoid using a skill builder if your agent architecture falls outside stateless, memory, or autonomous patterns, or if your repository does not align with BMad Method standards for capability routing and scaffolding.