How does it handle probability calibration guarantees?

The audit checks if the plugin delegates probability mathematics to VennAbers or IntervalRegressor reference logic, flagging any implementations that attempt to replace the core math wholesale.

Which metadata fields does this skill validate?

It validates fields like schema_version, name (reverse-DNS), version (SemVer), provider, capability tags, and modality-specific metadata introduced in ADR-033.

What is the primary goal of the CE Plugin Audit skill?

The primary goal is to ensure that plugins for the Calibrated Explanations library adhere to specific architectural contracts (ADRs) that guarantee calibration reliability and system stability.

What does the core/plugin boundary audit look for?

It checks for 'red flag' imports where a plugin tries to import implementation details from the core package rather than using approved protocols, dataclasses, or exceptions.

Does it check for performance issues?

Yes, Dimension 6 specifically audits for lazy imports, ensuring heavy dependencies like Matplotlib or Pandas are not imported at the top-level to keep the package responsive.

CE Plugin Audit

Name: CE Plugin Audit
Author: Moffran

byMoffran

•

Ciencia de Datos y ML

Audits Calibrated Explanations plugin implementations to ensure compliance with registry trust rules, metadata validity, and architectural contract standards.

The CE Plugin Audit skill provides a comprehensive framework for validating plugins within the Calibrated Explanations (CE) ecosystem. It enforces strict adherence to Architectural Decision Records (ADRs), verifying everything from metadata schemas and capability tags to internal protocol implementations and core-boundary isolation. By evaluating seven distinct audit dimensions, it ensures that plugins maintain calibration guarantees, handle fallbacks transparently, and optimize performance through lazy loading, providing developers with a structured report to ensure their extensions are registry-ready and maintainable.

Características Principales

01Audits fallback visibility and performance-critical lazy import patterns

02Verifies capability tags for classification, regression, and explanation types

0375 GitHub stars

04Enforces ADR-001 boundary rules to prevent leakage of core implementation details

05Validates plugin_meta against ADR-006 schema and reverse-DNS naming standards

06Checks Interval Calibrator Protocol compliance for VennAbers and probability logic

Casos de Uso

01Refactoring legacy plugins to meet modern ADR-033 modality and metadata standards

02Debugging non-conformant probability outputs or broken calibration guarantees

03Validating a new plugin implementation before submission to the CE registry

Características Principales

01Audits fallback visibility and performance-critical lazy import patterns

02Verifies capability tags for classification, regression, and explanation types

0375 GitHub stars

04Enforces ADR-001 boundary rules to prevent leakage of core implementation details

05Validates plugin_meta against ADR-006 schema and reverse-DNS naming standards

06Checks Interval Calibrator Protocol compliance for VennAbers and probability logic

Casos de Uso

01Refactoring legacy plugins to meet modern ADR-033 modality and metadata standards

02Debugging non-conformant probability outputs or broken calibration guarantees

03Validating a new plugin implementation before submission to the CE registry