RationaleAnchor: Preserving Architectural Intent Across Human and AI Transitions

Abstract

This paper introduces RationaleAnchor, a specification and reference implementation for preserving architectural intent as a first-class, verifiable artifact. RationaleAnchor defines a YAML schema capturing purpose, canonical goals, non-goals, invariants, dependencies, evidence, status, and human authority for every significant artifact. It provides automated validation of structural integrity, bidirectional dependency consistency, invariant verifiability, and succession plan executability. The reference implementation (Python 3.10+, zero mandatory dependencies) includes a CLI with 7 commands: validate, graph, check-drift, diff, intent-governance, conflict-check, succession-plan. RationaleAnchor extends Knowledge Continuity Engineering (KCE) with Layer 0 (Intent), resolving the bus factor of intent: when the original architect is unavailable, structured intent records preserve the rationale for design decisions. Validated on three artifacts from the BRIDGRAI ecosystem (KCE, CAMP, SIGNA) with 7/7 tests passing.