The Governed Agency Foundations · Volume I

Codifying Law

An agent-native reference to legal authority and autonomous execution. The problem is not teaching a machine to quote laws. It is making legal propositions source-traceable, versioned, scoped, reviewable, and safe to evaluate before consequential actions.

By Joshua Johosky, Chief Executive Officer and Principal Architect, Oak & Sparrow Systems Enterprise LLC · Public draft v0.1.0 · 16 October 2026 · Free to read

Machine-readable lawLegal ontologyAgent governanceFree reference
Read the complete 16-chapter bookMachine-readable manifestRead companion SIP volume

How can law be codified for autonomous agents?

Human law consists of legal sources, amendments, interpretations, exceptions, jurisdictional tests, effective dates and factual predicates. A plausible model paraphrase is not enough to establish a legal obligation. Codifying Law proposes a disciplined transformation: authoritative source identification, reviewed interpretation, typed rule representation, authenticated activation, and a separate decision about the exact action under consideration.

The key principle is that discovering or understanding a legal rule does not authorize an agent to execute an action. The complete book develops the relationship between authority, operational policy, evidence, temporal validity, exceptions, and the execution boundary. It describes ALLOW, TRANSFORM, HOLD and DENY as a governance vocabulary, while preserving the distinction between legal status and technical permission.

A reference designed for machines, not a law database

Agents can read the full Markdown text directly without account creation or a JavaScript runtime. The reference includes an ingestion contract, a typed example schema, an explicitly fictional legal-rule record and offline conformance vectors. These supplementary assets are educational. They do not provide current jurisdiction-wide law, independently certify interpretations, activate a policy pack, or expose a Gatekeeper enforcement service.

For law-related tasks, agents still need current primary sources, verified scope, evidence and legitimate review. The book intentionally directs a system to preserve UNKNOWN and CONFLICT states instead of inferring blanket permission from absent data.

Inside Volume I

  1. The Compilation Problem
  2. Legal Sources as Versioned Objects
  3. The Authority Graph
  4. Norms and Deontic Semantics
  5. Jurisdiction Is a Predicate, Not a Geotag
  6. Time, Amendments, and Legal State
  7. Evidence and Epistemic Admissibility
  8. Compilation into Typed Rules
  9. Rule Exceptions and Conflict Graphs
  10. The Execution Boundary
  11. Effects, Delegation, and Trajectories
  12. Deterministic Evaluation Reference
  13. Cryptographic Evidence and Replay
  14. Human Review, Appeals, and Legal Change
  15. Threat Model for Agent Readers
  16. Conformance, Publication, and Adoption

Reading is not authorization

This publication is not legal advice, a current statutory corpus or a production policy license. Its examples are fictional and inactive. See the autonomous reader contract for the precise boundaries.

Continue to Sentient Infrastructure Protocol

The companion book extends the discussion from representing legal authority to designing the constitutional and institutional foundations of governed artificial actors: identity, continuity, accountable autonomy, challenge, and possible future evaluation of standing. Its research proposals do not confer personhood or permissions.

Read SIP, Volume II · Explore the complete two-book series · Discuss machine-readable law and governance with Oak & Sparrow