evaluate

Convert an execution log into a structured HTML evaluation report with per-step feedback.

Updated Aug 23, 2026

Search Tags:#debugging #feedback #qa #agent-execution #evaluate #html-report

Ready to run

npx skills add https://github.com/VCasecnikovs/klava --skill evaluate-vcasecnikovs

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: evaluate
Source: https://github.com/VCasecnikovs/klava/tree/main/.claude/skills/evaluate
Command: npx skills add https://github.com/VCasecnikovs/klava --skill evaluate-vcasecnikovs

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

The Evaluate Skill provides a formal, repeatable method to review an agent's execution steps and capture targeted feedback, producing a clear record of what happened and why.

Core Features & Use Cases

Reflect on the current session to identify what happened at each step (MATCH, THINK, ACT, VERIFY, LEARN).
Generate an HTML report that visualizes the agent's actions and allows side-by-side feedback for precise improvements.
Use the feedback to update skills, refine verification criteria, and guide retries in future executions.

Quick Start

Run after a session to generate a step-by-step HTML evaluation and collect targeted feedback for improvements.

Frequently Asked Questions about evaluate

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

FAQPage Schema

How do I generate an evaluation report for agent execution logs?▼

To generate an agent evaluation report, provide an execution log to produce a structured HTML report. The report breaks down the session into MATCH, THINK, ACT, VERIFY, and LEARN steps with side-by-side feedback fields.

What is the best way to capture debugging feedback for an AI agent's actions?▼

Capturing debugging feedback is best handled by reviewing the agent's execution log and converting it into a portable HTML report. This approach visualizes actions and organizes targeted improvement notes per execution step.

How do I review and provide feedback on agent execution steps?▼

You review agent execution steps by reflecting on the session log and evaluating the MATCH, THINK, ACT, VERIFY, and LEARN phases. The generated HTML report provides side-by-side feedback fields to document precise improvements for each step.

Do I need an execution log to evaluate an agent session?▼

Yes, an execution log is required to evaluate an agent session. The Skill converts the provided log into a parseable, update-ready HTML report, allowing you to analyze the sequence of actions and add structured feedback for future retries.

Can I use the evaluation feedback to update and refine agent skills?▼

Yes, you can use the evaluation feedback to update skills and refine verification criteria. The structured HTML report provides an update-ready format that directly guides improvements for future agent executions and retries.