architecture-foundation

You can automate architecture binding by identifying concerns from Foundation capabilities and emitting per-concern binding contracts. The process wraps chosen GSD skills non-interactively to generate bound contracts and write a dependency DAG directly into capabilities.yaml.

Generating a dependency DAG from capabilities.yaml is best handled by expanding contract dependencies and writing them back into the existing depends_on field. This approach avoids creating a second DAG file while preserving acyclic graph validation and audit trails.

Architectural concerns are identified by partitioning Foundation capabilities into up to five concerns based on arch_surface or ADR-derived signals. When extra capabilities exist, they are folded deterministically into the existing concern partitions.

Yes, you can use this approach to write traceability data linking generated contracts back to original capabilities. It ensures deterministic output suitable for downstream SPEC and planning stages while preserving full audit trails across the dependency DAG.

Architecture binding should run immediately after the user stories phase and before downstream SPEC and planning stages. Operating at this point ensures the Foundation capabilities include arch_surface and ADR signals needed to fan out and generate accurate per-concern contracts.

DAG validation fails when expanded contract dependencies create cyclic graphs within capabilities.yaml. Ensure the architecture binding process maintains acyclic graphs by correctly mapping per-concern contract dependencies into the depends_on field without circular references.