Pattern Proposal: Typed Contracts for Agent Context

Summary

Pydantic and JSON Schema dependencies are used with dedicated model modules to define typed governance and steering contracts.

Candidate Pattern

• ID: typed-contracts-for-agent-context
• Classification: intelligence
• Maturity: strategic_direction
• Type: emerging

Observed Signals

No specific signals listed.

Evidence

Supporting Evidence

• pyproject.toml - Declares pydantic and jsonschema dependencies.
• src/cas_runner/governance/models.py (GovernanceAction, ComplianceFinding, GovernanceHealth, GovernanceRisks) - Defines typed governance context.
• src/cas_runner/steering/models.py (PromptFact, CapabilityRecord, CapabilitySet, ArchitecturalDelta) - Defines typed steering context.

Scope Assessment

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

Rationale

• Intent: Make shared agent context, governance data, and steering facts explicit, validatable, and inspectable.
• Context: Visible in governance models, steering models, and project dependencies for pydantic and jsonschema.
• Force: LLM-generated or agent-derived data must be verifiable.
• Force: Multiple components need a common understanding of actions, facts, and capability records.
• Force: Loose dictionaries would make governance and tracing unreliable.

Traceability

Artifact source: proposed-patterns