Library System

The unified library architecture that ensures consistency across all commands.

The Problem

Without a library system:

• Each command duplicates Phase 0 logic
• Updates require changing multiple files
• Inconsistencies creep in over time
• Maintenance becomes expensive

The Solution

┌─────────────────────────────────────────────┐
│                  Commands                    │
│  prompt.md  prompt-hybrid.md  prompt-tech.md │
│      │            │              │          │
│      └────────────┼──────────────┘          │
│                   ▼                          │
│  ┌─────────────────────────────────────────┐│
│  │     prompt-perfection-core.md           ││
│  │        (Canonical Phase 0)               ││
│  └─────────────────────────────────────────┘│
│                   │                          │
│      ┌───────────┼───────────┐              │
│      ▼           ▼           ▼              │
│  ┌────────┐ ┌────────┐ ┌────────┐          │
│  │Technical│ │Article │ │ Hybrid │          │
│  │ Adapter│ │ Adapter│ │ Adapter│          │
│  └────────┘ └────────┘ └────────┘          │
└─────────────────────────────────────────────┘

Core Library

Location: .claude/library/prompt-perfection-core.md

Contains:

• Universal Phase 0 flow (Steps 0.1-0.6)
• 6-criteria completeness check
• Approval gate pattern
• Optional Phase 0.15 (Predictive Intelligence)

Usage in Commands

## Phase 0: Prompt Perfection

**Import:** Use Phase 0 from @.claude/library/prompt-perfection-core.md
**Adaptation:** Technical (from @.claude/library/adapters/technical-adapter.md)

Benefits

1. Single source of truth - Phase 0 defined once
2. Consistency - All commands behave the same
3. Easy updates - Change once, all benefit
4. Smaller commands - Less duplication

Adapters

Adapters extend core library for specific domains.

Technical Adapter

Location: .claude/library/adapters/technical-adapter.md

Adds to completeness check:

• Technical Stack
• Architecture
• Code Location
• Testing Strategy

Adds question templates:

• Implementation method choices
• Framework-specific options
• Testing approach selection

Article Adapter

Location: .claude/library/adapters/article-adapter.md

Adds to completeness check:

• Target Audience
• Platform
• Tone and Style
• Content Length

Hybrid Adapter

Location: .claude/library/adapters/hybrid-adapter.md

Adds enhancements:

• Complexity detection engine
• Agent context gathering
• Multi-agent verification
• Learning system integration

Intelligence Components

Location: .claude/library/intelligence/

Predictive Intelligence Core

Implements Phase 0.15:

• Journey stage detection
• Domain risk analysis
• Pattern recognition
• Proactive warnings
• Next-steps prediction

Relationship Mapper

Connects work to previous sessions:

• Related decisions
• Similar implementations
• Dependent changes

Warning System

Proactive warnings by domain:

• Authentication risks
• Payment processing security
• Database concerns
• API vulnerabilities

Directory Structure

.claude/library/
├── prompt-perfection-core.md      # Canonical Phase 0
├── adapters/
│   ├── technical-adapter.md       # Technical domain
│   ├── article-adapter.md         # Content creation
│   ├── session-adapter.md         # Session management
│   └── hybrid-adapter.md          # Hybrid intelligence
├── intelligence/
│   ├── predictive-intelligence-core.md
│   ├── relationship-mapper.md
│   ├── warning-system.md
│   └── next-steps-predictor.md
├── orchestration/
│   ├── lead-agent-core.md         # Multi-agent coordination
│   ├── iteration-engine.md        # Research iterations
│   └── result-aggregator.md       # Combine agent results
└── agents/
    ├── explore-agent.md
    ├── pattern-agent.md
    ├── security-agent.md
    └── performance-agent.md

How Commands Reference Library

Using @ Import Syntax

Commands reference library using Claude Code's native @ syntax:

**Import:** @.claude/library/prompt-perfection-core.md

This tells Claude to load and follow the referenced file's instructions.

Layered Application

## Phase 0: Prompt Perfection

**Import:** Use Phase 0 from @.claude/library/prompt-perfection-core.md
**Adaptation:** Technical with Hybrid Intelligence
  - @.claude/library/adapters/technical-adapter.md
  - @.claude/library/adapters/hybrid-adapter.md
**Rules:** @.claude/rules/technical-patterns.md

Path-Specific Rules

Location: .claude/rules/

Rules apply based on file path using YAML frontmatter:

---
paths: .claude/commands/prompt-technical.md
---

# Technical Patterns

Rules for technical implementation analysis...

Available Rules

Rule File	Applies To
technical-patterns.md	Technical commands
command-conventions.md	All commands
library-standards.md	Library files
config-validation.md	Config files

Creating New Commands

To create a new command:

1. Create command file:

# /my-command

## Phase 0: Prompt Perfection

**Import:** @.claude/library/prompt-perfection-core.md
**Adaptation:** [None / Technical / Article / Custom]

[Command-specific phases...]

2. Create adapter if needed:

# My Domain Adapter

**Extends:** prompt-perfection-core.md

## Additional Completeness Criteria

- Custom Criterion 1
- Custom Criterion 2

3. Add to skills (optional):

{
  "name": "my-command",
  "command": "/my-command",
  "description": "...",
  "file": ".claude/commands/my-command.md"
}

Best Practices

For Library Changes

1. Test with multiple commands - Changes affect all users
2. Preserve backward compatibility - Don't break existing
3. Document changes - Update version history
4. Use adapters - Domain-specific logic belongs in adapters

For Command Authors

1. Reference, don't duplicate - Use @imports
2. Create adapters - For domain-specific needs
3. Follow conventions - Check existing commands
4. Test Phase 0 - Verify the flow works

Related

• Phase 0 - The core flow
• Hybrid Intelligence - Agent integration
• Configuration - Config files