domain-context

domain-context implements Progressive Domain Crystallization (PDC): a protocol for building and maintaining a living domain knowledge base for a custom business application, accumulated incrementally across sessions through human-AI collaboration. It solves the problem that an AI assistant does not know a company's internal terminology, entities, relationships, or business rules by default — and that this knowledge is rarely defined fully upfront but grows over time. The AI reads a structured domain file at the start of every session, uses the business's terminology exactly as defined, and flags gaps inline while it works.

The core protocol has three parts. At session start the AI locates and reads the domain file (reading a single file completely, or, for a split structure, reading the core index and pulling in only the relevant detail files via progressive disclosure). During work it uses defined entity names and terminology, and marks unknowns, unclear relationships, and inferred rules with explicit inline tags, never inventing domain facts. At session end it generates a 'Proposed Domain Updates' section — new entities, clarified terminology, relationships, business rules, and open questions — which the human reviews and manually applies, keeping the human as the domain authority and ensuring the knowledge base is never auto-corrupted. It ships a 14-section DOMAIN.md template with tier annotations, seed questions to bootstrap a new domain, an entity template, and an anti-patterns reference, plus a MAJOR.MINOR versioning scheme.

It targets teams building custom business applications who want the assistant to accumulate institutional knowledge — terminology, entities, state models, flows, stakeholders, and rules — across sessions. It is a pure documentation/knowledge-management skill requiring no system packages or network access, and is entirely benign.