At what points are guardrails enforced?

Guardrails are triggered during resource creation/updates, promotion requests or merges, and at the submission or start of a run execution.

How do I register a new contract kind?

Use the ContractRegistry.register method, providing a unique namespaced kind (e.g., 'signal.bounds'), a JSON schema, and a corresponding validator class.

What is the 'Law vs Police' architecture in OptAIC?

It is an architectural pattern where Definition resources (The Law) contain the validation contracts, while a central Guardrails Engine (The Police) reads and enforces them at specific gates without manual attachment.

What is the difference between BLOCK and WARN enforcement levels?

The BLOCK level prevents the action from completing (common in official subspaces), while WARN allows the action but logs issues (common in staging or development subspaces).

Where are validators typically implemented?

Validators should be implemented in the optaic/guardrails/validators/ directory and must be designed as pure, deterministic functions.

Guardrails Contracts for OptAIC

Name: Guardrails Contracts for OptAIC
Author: colingwuyu

bycolingwuyu

数据科学与机器学习

Standardizes the implementation of data validation, risk limits, and compliance contracts within the OptAIC framework.

关于

This skill equips Claude with specialized knowledge for designing and implementing validation contracts using the 'Law vs. Police' architecture. It guides developers through creating schema-driven guardrails for signal bounds, dataset integrity, and portfolio constraints, ensuring that validation rules are embedded directly into resource definitions. By following these patterns, developers can ensure that the Guardrails Engine automatically enforces data quality and compliance across various lifecycle gates, from resource creation to production promotion.

主要功能

Supports pure and deterministic validator implementation patterns
0 GitHub stars
Provides standardized patterns for signal bounds, dataset schemas, and portfolio constraints
Implements the 'Law vs. Police' architecture for centralized rule definition and automated enforcement
Guides the creation of namespaced contract kinds and JSON schema-based validation
Defines enforcement policies for multi-stage environments like staging and official subspaces

使用场景

Implementing complex portfolio constraints such as sum-to-one weights and leverage limits
Validating Point-In-Time (PIT) correctness and data freshness in financial datasets
Enforcing value range bounds on trading signals to prevent catastrophic outliers

关于

主要功能

Supports pure and deterministic validator implementation patterns
0 GitHub stars
Provides standardized patterns for signal bounds, dataset schemas, and portfolio constraints
Implements the 'Law vs. Police' architecture for centralized rule definition and automated enforcement
Guides the creation of namespaced contract kinds and JSON schema-based validation
Defines enforcement policies for multi-stage environments like staging and official subspaces

使用场景

Implementing complex portfolio constraints such as sum-to-one weights and leverage limits
Validating Point-In-Time (PIT) correctness and data freshness in financial datasets
Enforcing value range bounds on trading signals to prevent catastrophic outliers