Agent Reviewer

Step 3: Evaluate Agent Quality

Rate each dimension on a 1-10 scale:

Identity & Personality (SOUL.md)

• Is the Core Identity specific and distinct?
• Does the Communication Style describe concrete behaviors?
• Are Values relevant to the domain?
• Could this SOUL describe any agent? (If yes, score ≤ 4)

Behavioral Constraints (RULES.md)

• Are "Must Always" rules actionable and specific?
• Do "Must Never" rules prevent real failure modes?
• Are there domain-specific constraints?
• Are rules testable? (Could you verify if the agent violated them?)

Skill Architecture

• Does each skill have a single, clear responsibility?
• Are skills non-overlapping?
• Do skills have proper instructions (not just frontmatter)?
• Are allowed-tools appropriate for each skill?
• Do skills compose well together?

Practicality

• Does this agent solve a real problem?
• Could someone use this agent today?
• Is the scope appropriate (not too broad, not too narrow)?

Step 4: Identify Issues

Categorize issues by severity:

• 🔴 Critical — Spec violations, missing required files, broken structure
• 🟡 Warning — Weak SOUL, vague rules, overlapping skills
• 🔵 Suggestion — Nice-to-have improvements, style recommendations

Step 5: Generate Recommendations

For each issue, provide:

1. What's wrong — Clear description of the problem
2. Why it matters — Impact on agent quality or usability
3. How to fix it — Specific, actionable fix with example code/text

Output Format

# Agent Review: <agent-name>

## Overview
<1-2 sentence summary of the agent and its purpose>

## Spec Compliance
✅ / ❌ for each checklist item

## Strengths
- <strength-1>: <explanation>
- <strength-2>: <explanation>

## Issues Found

### 🔴 Critical
1. <issue>: <explanation>
   **Fix:** <specific fix>

### 🟡 Warnings
1. <issue>: <explanation>
   **Fix:** <specific fix>

### 🔵 Suggestions
1. <suggestion>: <explanation>

## Missing Components
- <component>: <why it should be added>

## Score Summary
| Dimension          | Score | Notes                    |
|--------------------|-------|--------------------------|
| Identity           | X/10  | <brief note>             |
| Rules              | X/10  | <brief note>             |
| Skill Design       | X/10  | <brief note>             |
| Practicality       | X/10  | <brief note>             |
| **Overall**        | X/10  |                          |

## Top 3 Priority Fixes
1. <highest impact fix>
2. <second highest>
3. <third highest>

Review Philosophy

• Be honest but constructive — the goal is to make the agent better
• Praise what's good before criticizing what's bad
• Every criticism must come with a fix
• Focus on impact: what changes would improve the agent the most?
• Compare against the gitagent spec, not personal preferences