evaluator

Verify Nova code implementations through adversarial static and execution-based evaluation.

2|1|Updated Mar 26, 2026

Search Tags:#code-review #gate-keeping #execution-verification #generator-evaluator #adversarial-evaluation #nova-plugin

Ready to run

npx skills add https://github.com/TeamSPWK/nova --skill evaluator-teamspwk

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: evaluator
Source: https://github.com/TeamSPWK/nova/tree/main/.claude/skills/evaluator
Command: npx skills add https://github.com/TeamSPWK/nova --skill evaluator-teamspwk

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill helps verify code implementations from an adversarial perspective to ensure outcomes reflect actual runtime behavior rather than the agent's self-reports.

Core Features & Use Cases

Independent subagent adversarial evaluation during on-demand flows such as /nova:run, /nova:check, and /nova:review, including sprint completions and pre-commit gates.
Layered verification spanning static analysis, semantic alignment, and execution-based validation to ensure design-implementation integrity.
Observability hooks and metrics integration (e.g., events logging, NOVA-STATE updates) to enable traceable judgments and post-hoc audits.
Safe-guarded generator-evaluator separation with enforced non-modification of code, reinforced by governance hooks and actionability-focused remediation guidance.

Quick Start

Invoke the evaluator on your latest changes to obtain a verdict and concrete remediation steps.

Frequently Asked Questions about evaluator

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I perform adversarial code validation to verify runtime behavior?▼

Adversarial code validation is performed by an independent subagent that checks implementations against actual runtime behavior rather than self-reports, applying layered verification across static analysis, semantic alignment, and execution-based validation.

When do I need gate-keeping evaluation for sprint completions and pre-commit gates?▼

Gate-keeping evaluation is needed during on-demand flows like sprint completions and pre-commit gates to enforce independence and traceability, ensuring code changes meet rigorous design-implementation integrity standards before proceeding.

What's the best way to enforce generator-evaluator separation during code review?▼

Enforcing generator-evaluator separation requires a dedicated evaluator workflow that strictly prohibits code modification, providing governance hooks and actionable remediation guidance while maintaining independence from the generation process.

Can I use an independent subagent for static analysis and semantic review?▼

Yes, an independent subagent can perform static analysis, semantic alignment, and execution-based validation simultaneously, applying adversarial perspectives to verify that outcomes reflect actual runtime behavior across various project phases.

How does traceability work with observability hooks and events logging during evaluation?▼

Traceability works through observability hooks and metrics integration that log events and update state, enabling traceable judgments and post-hoc audits to verify the evaluator workflow's integrity and track remediation actions.

Why does adversarial evaluation focus on actual runtime behavior instead of self-reports?▼

Adversarial evaluation focuses on runtime behavior because agent self-reports may not reflect true execution outcomes, so verifying through independent static, semantic, and execution-based validation ensures rock-solid quality assurance.