---

name: compound-plugin-analyzer description: Analyze Every Compound Engineering plugin and compare to our plugin for improvement suggestions argument-hint: "[optional: focus area e.g. 'agents', 'skills', 'patterns', 'commands']" model: claude-opus-4-5-20251101 context: fork agent: Explore allowed-tools:

• WebFetch
• WebSearch
• Read
• Grep
• Glob extended-thinking: true

---

Compound Plugin Analyzer

Thoroughly analyze the Every Compound Engineering plugin from GitHub and compare it to the PSD Claude Coding System plugin, generating prioritized improvement suggestions.

Source Information

Every Compound Engineering Plugin:

• Repository: https://github.com/EveryInc/compound-engineering-plugin
• Raw content base: https://raw.githubusercontent.com/EveryInc/compound-engineering-plugin/main/

Our Plugin:

• Location: /Users/hagelk/non-ic-code/psd-claude-coding-system/plugins/psd-claude-coding-system

Analysis Workflow

Phase 1: Fetch Every Plugin Structure

Use WebFetch to retrieve remote plugin content:

1. README and Version:

   • Fetch: https://raw.githubusercontent.com/EveryInc/compound-engineering-plugin/main/README.md
   • Extract version, feature list, architecture overview

2. Plugin Metadata:

   • Fetch: https://raw.githubusercontent.com/EveryInc/compound-engineering-plugin/main/plugins/compound-engineering/.claude-plugin/plugin.json
   • Extract version, commands list

3. CLAUDE.md Patterns:

   • Fetch: https://raw.githubusercontent.com/EveryInc/compound-engineering-plugin/main/plugins/compound-engineering/CLAUDE.md
   • Extract architectural patterns, model selection, workflow patterns

4. Agents Directory:

   • Use GitHub API or WebSearch to find agent list
   • Fetch key agents to understand patterns:
     • Agent-native-reviewer
     • Data-migration-expert
     • Deployment-verification-agent
     • Language-specific reviewers (kieran-rails, kieran-python, kieran-typescript)

5. Skills Directory:

   • Identify all skill directories
   • Fetch SKILL.md files from key skills:
     • git-worktree
     • gemini-imagegen
     • agent-browser
     • compound (knowledge compounding)

6. Commands:

   • Identify command naming patterns (namespace prefix like /workflows:)
   • Fetch command files to understand structure

Phase 2: Analyze Our Plugin Structure

Use Read, Grep, and Glob to analyze local plugin:

1. Agents Analysis:

   CopyGlob: plugins/psd-claude-coding-system/agents/*.md
   

   • Count total agents
   • Categorize by function (domain, quality, meta-learning, multi-LLM)
   • Extract unique capabilities (UX heuristics, telemetry, etc.)

2. Skills Analysis:

   CopyGlob: plugins/psd-claude-coding-system/skills/*/SKILL.md
   Glob: plugins/psd-claude-coding-system/skills/*.md
   

   • Count total skills
   • Identify skill patterns (directory vs single file)
   • Extract frontmatter patterns

3. Hooks Analysis:

   CopyRead: plugins/psd-claude-coding-system/hooks/hooks.json
   Glob: plugins/psd-claude-coding-system/scripts/*.sh
   

   • Document telemetry collection approach
   • Identify hook patterns

4. CLAUDE.md Patterns:

   CopyRead: CLAUDE.md
   

   • Extract our architectural patterns
   • Document model selection strategy
   • Note workflow patterns

Phase 3: Gap Analysis

Compare across these dimensions:

3.1 Agent Coverage

• List agents they have that we don't
• List agents we have that they don't
• Identify complementary capabilities

3.2 Skill Patterns

• Compare directory structure (SKILL.md + /scripts/ vs flat)
• Compare frontmatter fields
• Compare model selection (inherit vs explicit)

3.3 Command Organization

• Compare namespace strategies
• Compare command discovery patterns
• Compare argument handling

3.4 Safety Mechanisms

• Compare confirmation patterns
• Compare checklist approaches
• Compare error handling

3.5 Documentation Patterns

• Compare CLAUDE.md structure
• Compare inline documentation
• Compare example usage

3.6 Unique Strengths

Theirs:

• Agent-native architecture validation
• Data migration expertise
• Git worktree management
• Figma design sync
• Multi-language reviewers

Ours:

• 68-heuristic UX evaluation
• Multi-LLM council (GPT-5, Gemini 3)
• Meta-learning system (10 commands)
• Automatic telemetry hooks
• Pre-implementation security

Phase 4: Generate Improvement Report

Output the analysis in this format:

Copy# Compound Plugin Analysis Report

## Executive Summary
- Every Plugin Version: [detected]
- Our Plugin Version: [from plugin.json]
- Analysis Date: [current date]
- Focus Area: [if specified, else "Full Analysis"]

## Gap Analysis

### Agents We Should Add
| Priority | Agent Name | Purpose | Implementation Complexity | Notes |
|----------|------------|---------|---------------------------|-------|
| HIGH | ... | ... | Low/Medium/High | ... |
| MEDIUM | ... | ... | ... | ... |
| LOW | ... | ... | ... | ... |

### Skills We Should Add
| Priority | Skill Name | Purpose | Implementation Complexity | Notes |
|----------|------------|---------|---------------------------|-------|
| HIGH | ... | ... | Low/Medium/High | ... |
| MEDIUM | ... | ... | ... | ... |
| LOW | ... | ... | ... | ... |

### Patterns to Adopt
1. **Pattern Name**
   - Description: What it does
   - Their Implementation: How they do it
   - Our Adaptation: How we should implement
   - Files to Modify: Specific files

2. ...

### Our Unique Strengths (Keep/Enhance)
| Strength | Current Implementation | Enhancement Opportunity |
|----------|----------------------|------------------------|
| UX Specialist | 68 heuristics | ... |
| Multi-LLM Council | GPT-5, Gemini | ... |
| Meta-Learning | 10 commands | ... |
| Auto Telemetry | Hook-based | ... |

## Recommended Actions

### Immediate (This Week)
1. [Action item with specific files]
2. ...

### Short-Term (This Month)
1. ...

### Long-Term (This Quarter)
1. ...

## Implementation Roadmap

### Phase 1: Quick Wins (1-2 days each)
- ...

### Phase 2: Medium Effort (3-5 days each)
- ...

### Phase 3: Significant Features (1+ weeks each)
- ...

## Raw Data

### Every Plugin Stats
- Total Agents: X
- Total Skills: X
- Total Commands: X
- Unique Patterns: [list]

### Our Plugin Stats
- Total Agents: X
- Total Skills: X
- Total Commands: X
- Unique Patterns: [list]

Focus Area Handling

If an optional focus area argument is provided, narrow the analysis:

• agents: Deep dive on agent comparison only
• skills: Deep dive on skill patterns only
• patterns: Focus on architectural/workflow patterns
• commands: Focus on command organization and naming
• safety: Focus on safety mechanisms and confirmations

When focused, still provide context from other areas but prioritize depth over breadth.

Key URLs Reference

Copy# README
https://raw.githubusercontent.com/EveryInc/compound-engineering-plugin/main/README.md

# Plugin metadata
https://raw.githubusercontent.com/EveryInc/compound-engineering-plugin/main/plugins/compound-engineering/.claude-plugin/plugin.json

# CLAUDE.md
https://raw.githubusercontent.com/EveryInc/compound-engineering-plugin/main/plugins/compound-engineering/CLAUDE.md

# Sample agents (adjust paths based on actual structure)
https://raw.githubusercontent.com/EveryInc/compound-engineering-plugin/main/plugins/compound-engineering/agents/[agent-name].md

# Sample skills
https://raw.githubusercontent.com/EveryInc/compound-engineering-plugin/main/plugins/compound-engineering/skills/[skill-name]/SKILL.md

Execution Notes

1. Rate Limiting: Space WebFetch calls appropriately to avoid rate limiting
2. Error Handling: If a URL fails, note it and continue with available data
3. Caching: Results can be saved for comparison over time
4. Completeness: Prioritize getting accurate counts and patterns over fetching every file

Example Output Snippet

Copy### Agents We Should Add
| Priority | Agent Name | Purpose | Complexity | Notes |
|----------|------------|---------|------------|-------|
| HIGH | deployment-verification-agent | Go/No-Go deployment checklists | Medium | Complements our security-analyst |
| HIGH | data-migration-expert | Validates ID mappings in migrations | Medium | We lack migration-specific tooling |
| MEDIUM | agent-native-reviewer | Validates AI-native architecture | Low | Useful for our meta-learning work |
| MEDIUM | worktree-manager | Parallel git worktree management | High | Nice for complex feature work |
| LOW | figma-design-sync | Sync Figma designs to web | High | Only if we use Figma |