bmad-spec

Spec distillation converts raw meeting transcripts into a structured product specification by extracting a five-field SPEC kernel. This transforms messy intent into a clean machine contract ready for implementation and review.

Distilling mixed-source notes into a canonical spec requires extracting stable capability IDs and preserving load-bearing source content. The process routes overflow into companion files to keep the core specification organized and machine-readable.

You can update an existing specification without losing continuity because the process preserves stable capability IDs, assumptions, and open questions. This ensures spec revisions maintain historical context while integrating new intent.

During requirements distillation, long-form references and diagrams are routed into companion files. This companion routing moves load-bearing overflow out of the core SPEC kernel so detailed content stays organized and accessible for downstream work.

Spec distillation works for PRDs, RFCs, briefs, and transcripts by applying preservation validation to generate a canonical SPEC kernel. It handles diverse input formats across both interactive and headless workflows uniformly.

You should use companion artifacts instead of a single spec file when load-bearing source content overflows the core kernel. Moving detailed references and long-form content into companion files prevents bloat while preserving the original intent.