Eval Skill - Prompt Evaluation Workflow

This skill guides you through evaluating Deca's prompts using the eval system.

Overview

The eval system tests whether the Agent behaves according to its prompts (IDENTITY.md, SOUL.md, etc.).

Key Constraint: LLM (you) judges the results manually. Scripts do NOT use LLM.

Workflow

[1. Run Cases] → [2. Judge Results] → [3. Generate Report]
    (script)        (LLM manual)         (script)