judge-evaluate

Evaluate patch.diff and handoff.json against task scope, generating verdict.json.

Updated Feb 10, 2026
One-click install
npx skills add https://github.com/OilProducts/agent-skills --skill judge-evaluate
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: judge-evaluate
Source: https://github.com/OilProducts/agent-skills/tree/main/judge-evaluate
Command: npx skills add https://github.com/OilProducts/agent-skills --skill judge-evaluate

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill automates the evaluation of proposed code changes and their accompanying artifacts, ensuring they meet task scope and verification checks, and producing a machine-readable verdict.

Core Features & Use Cases

  • Automated Code Review: Evaluate patch.diff and handoff.json against predefined checks.
  • Machine-Readable Verdicts: Generates verdict.json with clear pass, reject, or needs-human statuses.
  • Deterministic Evaluation: Executes specified commands (e.g., pytest, ruff check) for objective assessment.
  • Use Case: After a developer submits a code change, this Skill can automatically run tests and linters, then provide a structured judgment on whether the change is acceptable, needs rejection, or requires human intervention, all without manual review.

Quick Start

Use the judge-evaluate skill to evaluate the submitted patch and handoff artifacts for task 'TASK-0042'.

Frequently Asked Questions about judge-evaluate

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

FAQPage Schema
How do I automate code review for patch submissions without manual editing?

Automated code review evaluates patch.diff and handoff.json against predefined verification checks, producing a machine-readable verdict.json. It validates handoff integrity and runs deterministic evaluation commands like pytest or ruff check to accept, reject, or escalate code changes objectively.

What is a machine-readable verdict format for patch evaluation?

A machine-readable verdict for patch evaluation is a structured verdict.json file that outputs clear pass, reject, or needs-human statuses. It synthesizes evidence from deterministic evaluation commands and handoff artifact validation without modifying source files.

How do I run deterministic evaluation commands on a proposed implementation patch?

Deterministic evaluation commands run on a proposed implementation patch by executing specified checks like pytest and ruff check within the evaluation workflow. The skill validates handoff integrity first, then runs these commands to synthesize an objective verdict.json.

Can I use automated patch evaluation for tasks requiring a separate judging role?

Automated patch evaluation supports a separate judging role by accepting patch.diff and handoff.json from a doer, validating task scope and verification checks, and outputting a verdict.json. The judge can accept, reject, or escalate without editing source files directly.

What are the limitations of automated code change evaluation?

Automated code change evaluation does not edit source files and relies on deterministic evaluation commands. It produces needs-human statuses when checks fail or handoff integrity is compromised, meaning it cannot resolve ambiguous patches autonomously and requires human intervention for escalated cases.