eval-finalize

Aggregate turn-evaluation data into a 10-section step3_finalization.md document.

Updated Aug 23, 2026
One-click install
npx skills add https://github.com/MinhOmega/marlin-skill --skill eval-finalize
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: eval-finalize
Source: https://github.com/MinhOmega/marlin-skill/tree/main/.claude/skills/eval-finalize
Command: npx skills add https://github.com/MinhOmega/marlin-skill --skill eval-finalize

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It automates the creation of a structured Step 3 finalization report after completing multiple turns of evaluation, saving time and reducing manual synthesis errors.

Core Features & Use Cases

  • Generates a 10-section finalization document including trajectory, multi-axis ratings, and submission readiness.
  • Integrates data from all turns and the accepted_baseline.json to justify decisions with evidence.
  • Produces a clear, justified narrative suitable for submission reviews and handoffs.

Quick Start

Run the eval-finalize skill with a workspace path to generate step3_finalization.md after completing all turns; ensure step1_spec.md and accepted_baseline.json are present.

Frequently Asked Questions about eval-finalize

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

FAQPage Schema
What is a Step 3 finalization document for multi-turn evaluations?

A multi-turn evaluation finalization report is a structured 10-section document that summarizes trajectory, multi-axis ratings, and submission readiness for tasks with 3+ turns. It aggregates turn-evaluation data to justify decisions with evidence from accepted_baseline.json.

Can I automate finalization reporting for multi-turn trajectory evaluations?

To create step3_finalization.md, your workspace must contain completed turns, step1_spec.md, and accepted_baseline.json. These inputs provide the evaluation trajectory and baseline data needed to justify decisions with citations to previous turns.

How do I include citations to previous turns in a finalization report?

No, eval-finalize applies specifically to multi-turn evaluation tasks with 3 or more turns. It requires access to a workspace with completed turns to aggregate the trajectory data and generate citations to previous turns in the finalization document.

What are the limitations of automated finalization reporting for multi-turn evaluations?

Submission readiness is justified by integrating data from all completed turns and accepted_baseline.json to produce a clear, written narrative. The finalization document aggregates multi-axis ratings and trajectory evidence to support the submission review.