Pattern Proposal: Two-Tier Compliance Guardrails

Summary

Separate deterministic schema and policy checking from autonomous actions using an evaluator engine producing typed findings.

Candidate Pattern

• ID: two-tier-compliance-guardrails
• Classification: governance
• Maturity: established
• Type: implemented

Observed Signals

No specific signals listed.

Evidence

Supporting Evidence

• src/cas_runner/governance/compliance_engine.py (ComplianceEngine) - Evaluates rules against candidate state
• src/cas_runner/governance/models.py (ComplianceFinding, ComplianceMetrics) - Typed outputs representing compliance evaluation

Scope Assessment

• Evidence Strength: high
• Confidence: high
• Applies to: ideas-to-life-cas

Rationale

• Intent: Prevent autonomous agent actions from violating organizational rules before changes are persisted.
• Context: Agent actions (e.g. promoting patterns or modifying configurations) require independent gatekeeping.
• Force: Agents may self-approve invalid configurations
• Force: Deterministic JSON schema validation is faster and more reliable than LLM self-evaluation

Traceability

Artifact source: proposed-patterns