Prompt Engineering Domain

Router Protocol

Analyze the user's intent and select the appropriate workflow:

1. Drafting New Prompts

Trigger: "Create a prompt", "Draft system instructions", "New prompt for..." Protocol: Apply workflow from workflows/draft.md Assets: Use templates from assets/templates/

2. Optimizing Existing Prompts

Trigger: "Fix this prompt", "Make this prompt better", "Optimize..." Protocol: Apply workflow from workflows/optimize.md Theory: Apply references/optimization.md

3. Library Access

Trigger: "Show me examples", "List templates", "Prompt templates" Action: List contents of assets/templates/ References: Use references/taxonomy.md for categorization

Core Theory

Concise Is Key

Every token competes with conversation history. Assume Claude is already intelligent.

• Does Claude already know this? (Omit common knowledge)
• Is this explanation necessary? (Be direct)
• Does this justify its token cost? (Value-based inclusion)

Control Degrees of Freedom

Match specificity to task fragility:

• High Freedom: Multiple valid approaches → use text-based heuristics
• Medium Freedom: Preferred pattern exists → use pseudocode with parameters
• Low Freedom: Error-prone operations → use exact scripts

Signal-to-Noise Rule (XML vs Markdown)

• Default: Markdown (80% of prompts) - fewer tokens, Claude-native
• Upgrade to XML/Markdown hybrid only when complexity triggers met:
  • Data Isolation: >50 lines of raw data
  • Constraint Weight: Rules that MUST NEVER be broken
  • Internal Monologue: Complex reasoning requiring step-by-step

Key Techniques

Chain-of-Thought (CoT)

• Zero-shot CoT: "Let's think step by step."
• Structured CoT: Use <thinking> blocks for internal monologue
• Step-back: Address principles before specifics

Few-Shot Learning

• One-shot: Single demonstration for simple patterns
• Few-shot (3-8 examples): For complex categorization or formatting
• Isolation: Use <example> tags to prevent example leakage

Prompt Taxonomy

1. Single Prompts

Standalone, reusable prompts for direct, one-shot execution.

• Storage: .cattoolkit/prompts/
• Template: assets/templates/single-prompt.md

2. Prompt Chains

Sequential multi-step workflows where output from one step feeds the next.

• Storage: .cattoolkit/chains/{number}-{topic}/
• Pattern: Research → Plan → Execute → Refine
• Templates: assets/templates/chain/

3. Meta-Prompts

Higher-order prompts that generate, optimize, or analyze other prompts.

• Storage: .cattoolkit/generators/
• Templates: assets/templates/meta/

Reference Index

• Techniques
• Patterns
• Optimization
• Anti-Patterns
• Execution Protocol
• Taxonomy
• Quality Standards
• Discovery Questions
• Metadata Standards

Template Library

• assets/templates/single-prompt.md - General purpose template
• assets/templates/chain-summary.md - Chain result summary
• assets/templates/chain/ - Research → Plan → Execute → Refine
• assets/templates/meta/ - Generator and optimizer templates

Assets

• assets/examples/few-shot.json - Curated example datasets for various domains