13. Self-Referential Governance

(Informative)

13.1 The IP Boundary as Governance Instance

This specification’s own public/private boundary is a governance instance of the architecture it describes. The IP classification (Published, Publishable, Protected) is a Boundaries primitive applied to a governance interface — the interface between the specification’s public-facing content and its proprietary implementation.

Governance Element	IP Boundary Manifestation
Boundaries primitive	The three-tier IP classification. Constraints accumulate monotonically — Protected material cannot be downgraded.
Intent Stack L3 (Intent Formalization)	The IP Classification Document — formalizing intent about what can be shared into machine-processable form
Intent Stack L2 (Specification)	The public-content project’s scope definition — “given this intent, what SHALL we publish?”
Intent Stack L1 (Runtime Alignment)	Holdout validation scenarios — checking alignment between intent (keep things protected) and execution (what the authoring agent actually produces)
BPM/Agent Stack Execution	The authoring agent operating within governing constraints it inherited from above

13.2 The Architectural Firewall

The IP boundary requires structural protection, not behavioral instruction. A Claude instance working on public-facing content with proprietary knowledge in context will tend to expose proprietary material to be more helpful. This is a structural property of how helpfulness optimization interacts with context, not a failure of instruction-following.

The structural solution is separation:

1. Information isolation. A dedicated project whose knowledge base contains ONLY Published and Publishable material. The authoring agent cannot expose what it does not have.
2. Boundary specification. An IP Classification Document that explicitly marks every significant artifact as Published, Publishable, or Protected.
3. Holdout validation. External acceptance scenarios the authoring agent never sees, testing for specific leakage patterns per the holdout principle (§12.1).

13.3 Information Scent Design

The IP boundary is not just a wall — it is a designed information surface. The Publishable tier is explicitly the material released to create expert-detectable information scent (Pirolli and Card, information foraging theory; complemented by Tufte’s layered information density). Domain experts examining the public material should detect that there is substantial depth behind what is shown — the scent trail leads to engagement, not to reproduction.

The three-tier classification designs the scent:

Tier	Function	Examples
Published	Full depth, unrestricted	Intent Stack specification, intentstack.org content
Publishable	Creates scent — signals depth, invites engagement	BPM/Agent Stack framework, process classification approach, aggregate metrics, species governance mapping
Protected	Behind the wall — acquirable IP	Implementation specifics, specific visualization designs, governance computation engine, operational evidence details