bmad-agent-builder

To package an agent idea into a ready-to-use skill, this tool uses conversational discovery to define your agent's identity and capabilities, then outputs a self-contained unit with a root SKILL.md frontmatter and optional resource directories. This structured packaging ensures your agent is prepared for integration and governance.

The best way to structure agent skills for discovery and governance is to package them as self-contained units with consistent metadata. This includes a YAML frontmatter with a name and description, plus aligned optional directories like scripts, references, and assets for bootstrapping templates.

Yes, you can build a memory-enabled agent skill using guided conversational discovery. The process helps you define the agent's persona, creed, bond, and memory capabilities, creating a complete agent skeleton ready for integration with headless execution support.

Yes, you need pyyaml installed to validate a SKILL.md frontmatter. The validation process verifies that the YAML frontmatter includes the required name and description fields, and checks that any optional directories align with the bootstrapping templates.

When assembling an agent skill skeleton, no directories are strictly required, but optional directories like scripts, references, and assets are verified if they exist. The only strict requirement is a root SKILL.md file containing valid YAML frontmatter with a name and description.

Conversational discovery helps architect an agent skill by guiding you through defining the agent's persona, capabilities, memory, and behavior without getting lost in procedural details. It builds a complete agent skeleton including a creed, bond, and optional PULSE features ready for integration.