ObsidianWall Vocabulary

Implementation-specific terminology for ObsidianWall — Verdict, Sentinel, Compass, Forge, and the Two-Plane Doctrine.

This page defines vocabulary specific to the ObsidianWall implementation.

These are implementation terms, not category terms. For the foundational Programmable Assurance vocabulary — the Alignment Principle, Intent-Reality Gap, Governance Fragmentation, the Four Principles, and other discipline-level concepts — see programmableassurance.org/vocabulary.

Two-Plane Doctrine

The architectural separation of enforcement authority from advisory intelligence.

Deterministic Governance Plane — Where enforcement decisions live. Verdict, Sentinel, Risk Ledger. Every output is deterministic, reproducible, and attributable to a human-authored policy. AI may not govern here.

Probabilistic Intelligence Plane — Where advisory functions live. Compass, Governance Intelligence, Governance Economics. Every output informs decisions but never makes them.

AI may advise. AI may not authoritatively govern.

Verdict

The pre-deployment governance decision engine. ObsidianWall’s first executable. Evaluates infrastructure plans against policies and produces a typed governance decision before deployment executes.

Sentinel

The post-deployment governance observation engine. Verifies that operational reality stayed aligned with what the governance decision authorized.

Forge

The governance authoring engine. Unique: lives in both planes. Translates leadership intent into executable governance policy through a human-gated AI layer. The human is the bridge between planes.

Compass

The governance intelligence and economics engine. Analyzes outcomes across governance decisions. Closes the Assurance Feedback Loop.

Risk Ledger

Cryptographically signed, tamper-evident record of every governance decision, override, and risk acceptance. The fiduciary shield for security leaders.

Assurance Artifact

The immutable JSON record produced by every Verdict evaluation. Contains decision, conditions, trace, risk summary, notification manifest, and full governance reasoning chain.

Governance Decision

One of five typed outcomes produced by Verdict:

  • ALLOW — All conditions passed. Deployment authorized.
  • ALLOW_WITH_NOTIFICATION — Conditions passed. Stakeholder notification required.
  • ALLOW_WITH_APPROVAL_REQUIRED — Conditions passed. Formal approval required before proceeding.
  • DENY_WITH_OVERRIDE — Conditions failed. An authorized role may override.
  • DENY — Conditions failed. No override path. Hard block.

This vocabulary describes ObsidianWall specifically. For the Programmable Assurance discipline vocabulary, see programmableassurance.org/vocabulary.