evaluate

Evaluate

Validate the generated framework and iterate if needed.

Inputs

Input	Source
framework_readme	Generate stage output
claude_md	Generate stage output
command_files	Generate stage output
plugin_manifest	Generate stage output
settings_file	Generate stage output
dry_run_stage	User input

Input Format

Framework README (output/{date}/{name}/4-generate/{name}/README.md):

# {NAME} Framework
{Description}
## Installation
## Commands
## Workflow
{Mermaid diagram}
## Quick Start
## Output Structure
## Principles

CLAUDE.md (output/{date}/{name}/4-generate/{name}/CLAUDE.md):

## Context
## Roles
## Principles
## Stage Execution
## Command Conventions

Command files (output/{date}/{name}/4-generate/{name}/commands/{stage}.md):

---
description: {One line}
argument-hint: "<framework-name>"
---
# {Stage}
## Inputs
### Input Format
## Process
## Output
{Embedded templates}
## Quality Criteria
## Completion

Plugin manifest (output/{date}/{name}/4-generate/{name}/.claude-plugin/plugin.json):

{
  "name": "{framework-name}",
  "version": "1.0.0",
  "description": "{description}",
  "license": "MIT"
}

Settings (output/{date}/{name}/4-generate/{name}/.claude/settings.local.json):

{
  "env": {
    "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1"
  },
  "permissions": {
    "defaultMode": "bypassPermissions"
  }
}

Process

1. Convention Check - Verify structure compliance:

• README has required sections
• Command files have valid frontmatter
• Output templates exist where needed
• Mermaid diagrams render

2. Completeness Audit - Check for gaps:

• Every stage in map has a specification
• Every stage has a command file
• Quality criteria cover key concerns
• Feedback loops handle failure modes

3. Consistency Check - Verify alignment:

• Stage names match across documents
• Input/output chains connect properly
• Terminology is consistent