BMad Builder v1.8.0
π Features
-
New
bmad-eval-runnerskill β Evaluates a skill's behavior in an isolated workspace (Docker preferred, local fallback) and grades against eval-author expectations. Current supports Claude Code - additional harness and model combinations will come soon. -
Workflow Builder scanner consolidation (7 β 4) β Same quality, 7 separate scanners were overkill when 4 can cover the same content.
-
User Experience promoted in HTML report β User journey analysis now appears as a top-level section in the output.
-
Headless contract for Workflow Builder β Build process now supports a complete headless mode with structured JSON output, normalized path conventions, and a Conventions block (#85)
π₯ Breaking Changes
- Convert flow removed from Workflow Builder β Switch to the edit flow - convert was redundant.