Memoriant Governance Compiler Skill

A Claude Code plugin that transforms natural-language policy documents into executable governance graphs. Compile Markdown policies into typed DAGs, detect policy conflicts, trace compliance paths for specific scenarios, and generate audit evidence packages.

No servers. No Docker. Just install and use.

Install

/install NathanMaine/memoriant-governance-compiler-skill

Cross-Platform Support

This plugin works with multiple AI coding assistants:

Claude Code (Primary)

/install NathanMaine/memoriant-governance-compiler-skill

OpenAI Codex CLI

git clone https://github.com/NathanMaine/memoriant-governance-compiler-skill.git ~/.codex/skills/governance-compiler
codex --enable skills

Gemini CLI

gemini extensions install https://github.com/NathanMaine/memoriant-governance-compiler-skill.git --consent

Skills

Skill	Command	What It Does
Compile Policy	/compile-policy	Parse Markdown policy → typed DAG (JSON + DOT format)
Analyze Graph	/analyze-graph	Structural analysis: density, depth, orphaned nodes, coverage gaps
Detect Conflicts	/detect-conflicts	Find contradictory rules, duplicate mandates, frequency conflicts
Trace Compliance	/trace-compliance	Evaluate a scenario — PERMITTED / PROHIBITED / CONDITIONAL / AMBIGUOUS
Generate Audit	/generate-audit	Full audit evidence package: inventory, coverage matrix, conflict register

Agents

Agent	Best Model	Specialty
Policy Compiler	Opus 4.6 + extended thinking	Deterministic Markdown → DAG compilation
Compliance Evaluator	Opus 4.6 + extended thinking	Scenario evaluation, evidence-backed verdicts

Quick Start

# Compile a policy document
/compile-policy

# Find conflicts in the compiled graph
/detect-conflicts

# Evaluate a specific action
/trace-compliance "a service account reads production data without logging"

# Generate a compliance audit package
/generate-audit

Example: Policy Compilation

Input: Markdown policy document with sections and bullet rules

Output:

{
  "nodes": [
    {"id": "section-1", "label": "Data Handling", "type": "section", "level": 1},
    {"id": "rule-1", "label": "Encrypt data in transit", "type": "rule", "parent": "section-1"}
  ],
  "edges": [
    {"from": "section-1", "to": "rule-1", "label": "contains"}
  ]
}

Visualize the DOT output at GraphvizOnline.

Example: Compliance Trace

Scenario: "Agent spawns a sub-agent to execute a financial transaction"

Verdict: CONDITIONAL

Rules applied:
  [REQUIRES] rule-34: "Level 3: Irreversible actions require human supervisor approval"
  [REQUIRES] rule-4:  "Preserve audit trail for all decisions"
  [PERMITS]  rule-20: "Orchestrator agent holds delegation authority"

Conditions for PERMITTED:
  1. Human supervisor approves the transaction
  2. Audit trail entry is created

Use Cases

• AI agent governance: compile agent constitutions into enforceable graphs
• Compliance workflows: trace CMMC, HIPAA, or SOC 2 control applicability
• Policy management: detect conflicting rules before deployment
• Audit preparation: generate structured evidence for compliance reviews
• Multi-agent coordination: define and enforce agent behavior policies

Source Repository

Built from NathanMaine/governance-graph-compiler.

License

MIT — see LICENSE for details.