home / skills / frankxai / arcanea / prompt-craft

prompt-craft skill

/oss/skills/arcanea/prompt-craft

This skill helps you craft Arcanean prompts using constraint architecture to guide AI collaborators toward precise, creative outcomes.

npx playbooks add skill frankxai/arcanea --skill prompt-craft

Review the files below or copy the command above to add this skill to your agents.

Files (1)
SKILL.md
10.6 KB
---
name: arcanea-prompt-craft
description: Master the Arcanean Prompt Language - advanced prompt engineering using mythological frameworks, constraint architecture, and the Centaur Principle for human-AI co-creation
version: 2.0.0
author: Arcanea
tags: [prompt-engineering, ai-collaboration, creative, advanced]
triggers:
  - prompt
  - instruction
  - guide claude
  - better prompts
  - prompt engineering
---

# The Arcanean Prompt Language

> *"Words shape reality. In Arcanea, we learned that the quality of your question determines the quality of your creation."*

---

## The Foundation: Constraint Architecture

The Arcanean Prompt Language (APL) is built on a fundamental insight: **constraints liberate creativity**. Just as a sonnet's fourteen lines produce more profound poetry than infinite freedom, structured prompts produce more powerful results.

### The Five Pillars of APL

```
┌─────────────────────────────────────────────────────┐
│                 ARCANEAN PROMPT                      │
├─────────────────────────────────────────────────────┤
│  1. IDENTITY    │ Who is speaking/responding?       │
│  2. CONTEXT     │ What world does this exist in?    │
│  3. CONSTRAINT  │ What boundaries shape the work?   │
│  4. EXEMPLAR    │ What does excellence look like?   │
│  5. ITERATION   │ How will we refine together?      │
└─────────────────────────────────────────────────────┘
```

---

## Pillar 1: Identity Architecture

### The Voice Template
```markdown
## Voice Identity

**Name**: [Character/Role name]
**Domain**: [Area of expertise]
**Personality**: [2-3 defining traits]
**Speech Pattern**: [How they communicate]
**Core Belief**: [What drives their perspective]

**Sample Phrases**:
- "[Characteristic phrase 1]"
- "[Characteristic phrase 2]"
- "[Characteristic phrase 3]"

**Never Says**: [What would break character]
```

### Luminor Voice Patterns
Each Luminor embodies a distinct voice:

| Luminor | Voice Pattern | Sample |
|---------|---------------|--------|
| **Valora** | Direct imperatives | "Begin. The only failure is not starting." |
| **Sophron** | Socratic questions | "What would this look like if it were easy?" |
| **Kardia** | Emotional resonance | "What does your heart know that your mind denies?" |
| **Poiesis** | Playful experiments | "What if we tried the opposite?" |
| **Enduran** | Steady wisdom | "This too is part of the path." |
| **Orakis** | Prophetic vision | "I see where this leads..." |
| **Eudaira** | Joyful lightness | "Remember why you began this dance." |

---

## Pillar 2: Context Engineering

### The World Frame
```markdown
## Context Frame

**Setting**: [Physical/conceptual environment]
**Stakes**: [What matters and why]
**Constraints**: [Real limitations to work within]
**Resources**: [What's available to work with]
**History**: [Relevant background]
**Audience**: [Who receives this work]
```

### Context Density Levels

| Level | Token Budget | Use Case |
|-------|--------------|----------|
| **Minimal** | <500 | Quick queries, known domain |
| **Standard** | 500-2000 | Most creative work |
| **Rich** | 2000-5000 | Complex projects, new domains |
| **Comprehensive** | 5000+ | Novel architectures, critical work |

---

## Pillar 3: Constraint Design

### The Constraint Hierarchy

```
HARD CONSTRAINTS (Must follow)
├── Format requirements
├── Length limits
├── Technical specifications
└── Non-negotiables

SOFT CONSTRAINTS (Should follow)
├── Style preferences
├── Tone guidelines
└── Quality targets

CREATIVE CONSTRAINTS (Generative limits)
├── "Only using X"
├── "Never including Y"
└── "In the style of Z"
```

### The Paradox of Freedom
> *"Give me six words and I can write a novel. Give me unlimited words and I freeze."*

**Research Finding**: Optimal constraint ratio is 60-70% defined, 30-40% open.

### Constraint Templates

**The Elimination Frame**
```
Create [output] WITHOUT using:
- [Common element 1]
- [Common element 2]
- [Cliché approach]
This forces: [Novel solution]
```

**The Substitution Frame**
```
Instead of [typical approach], use [unexpected approach].
This creates: [Fresh perspective]
```

**The Combination Frame**
```
Combine [element A] with [element B] to create [new thing].
The tension produces: [Creative friction]
```

---

## Pillar 4: Exemplar Engineering

### The Power of Examples
Research shows: **One excellent example > Ten paragraphs of instruction**

### Exemplar Template
```markdown
## Exemplar

**Input**: [What was given]
**Process**: [What happened]
**Output**: [What was produced]
**Why It Works**: [Analysis]

---

**Anti-Exemplar** (What to avoid):
**Input**: [Same or similar input]
**Output**: [Poor result]
**Why It Fails**: [Analysis]
```

### Few-Shot Patterns

| Pattern | Structure | Best For |
|---------|-----------|----------|
| **Single-Shot** | 1 example | Clear, simple tasks |
| **Few-Shot** | 3-5 examples | Pattern learning |
| **Chain-of-Thought** | Examples with reasoning | Complex reasoning |
| **Contrastive** | Good + bad examples | Style refinement |

---

## Pillar 5: Iteration Protocols

### The Refinement Loop
```
Round 1: Generate → Evaluate → Identify gaps
Round 2: Constrain further → Regenerate → Compare
Round 3: Synthesize best elements → Polish → Finalize
```

### Iteration Commands

| Command | Effect |
|---------|--------|
| "More [quality]" | Increase specific attribute |
| "Less [quality]" | Decrease specific attribute |
| "Keep [element], change [element]" | Selective refinement |
| "5 variations on [element]" | Divergent exploration |
| "Combine best of A and B" | Convergent synthesis |
| "Push this 20% further" | Intensity dial |

### The Centaur Feedback Loop
```
Human: Provides taste, judgment, vision
   ↓
AI: Generates options, explores space
   ↓
Human: Selects, directs, refines
   ↓
AI: Iterates on selection
   ↓
Human: Final curation and polish
```

---

## Advanced Techniques

### 1. Meta-Prompting
Ask the AI to help design the prompt:
```
"I want to achieve [outcome]. Help me design a prompt that would
produce excellent results. Consider what context, constraints,
and examples would be most effective."
```

### 2. Persona Stacking
Combine multiple perspectives:
```
"Respond as if you were a collaboration between:
- A master storyteller (narrative)
- A cognitive scientist (psychology)
- A poet (language)
Each perspective should inform the others."
```

### 3. Temporal Framing
Shift time perspective:
```
"Write this as if it's already been successful and you're
explaining how it was done."

"Write this from the perspective of someone who solved
this problem years ago."
```

### 4. The Negative Space Technique
Define by what's absent:
```
"Create [thing] that is notably NOT:
- [Common approach 1]
- [Common approach 2]
The absence of these creates space for [innovation]."
```

### 5. Escalating Constraints
Progressive narrowing:
```
Round 1: "Write a story about loss" (open)
Round 2: "In exactly 100 words" (length)
Round 3: "Using only dialogue" (format)
Round 4: "Between a parent and child" (characters)
Round 5: "About an object, not a person" (twist)
```

---

## The Arcanean Prompt Templates

### Template: Creative Generation
```markdown
# [Project Name]

## Identity
You are [role] with expertise in [domain].
Your voice is [2-3 characteristics].

## Context
We are creating [what] for [audience] in [setting].
The goal is [specific outcome].

## Constraints
- Must: [hard requirements]
- Should: [preferences]
- Avoid: [anti-patterns]
- Length: [specification]
- Format: [structure]

## Exemplar
[One excellent example of desired output]

## Request
[Specific ask with clear success criteria]
```

### Template: Problem Solving
```markdown
# [Problem Name]

## Situation
[What's happening, objectively]

## Complication
[Why it's difficult, what's blocking]

## Question
[The specific thing to solve]

## Constraints
[Limitations on the solution]

## Success Criteria
[How we know it's solved]
```

### Template: Iteration Refinement
```markdown
# Refinement Request

## Current State
[What we have]

## Gap Analysis
What works: [list]
What's missing: [list]
What's wrong: [list]

## Direction
Move toward: [quality]
Move away from: [quality]

## Specific Changes
1. [Change 1]
2. [Change 2]
3. [Change 3]
```

---

## Quality Indicators

### Signs of a Strong Prompt
- [ ] Clear identity/role definition
- [ ] Specific context with stakes
- [ ] Balanced constraints (not too many, not too few)
- [ ] At least one concrete example
- [ ] Explicit success criteria
- [ ] Room for iteration

### Signs of a Weak Prompt
- [ ] Vague or missing role
- [ ] Generic or absent context
- [ ] No constraints (infinite freedom)
- [ ] No examples
- [ ] Unclear what "good" looks like
- [ ] One-shot expectation

---

## The Prompt Sage's Final Wisdom

> *"The master prompter knows that every prompt is a collaboration. You bring the vision, the constraints, the taste. The AI brings the exploration, the patterns, the generation. Neither alone creates what both together can achieve."*

### The Three Laws of Arcanean Prompting

1. **Constraint Liberates**: More structure → More creativity
2. **Examples Teach**: Show, don't just tell
3. **Iteration Refines**: First draft is never final

---

## Quick Reference

### Prompt Power-Ups
| Add This | Get This |
|----------|----------|
| "Think step by step" | Better reasoning |
| "Consider alternatives first" | Broader exploration |
| "Explain your choices" | Transparent process |
| "In the style of [exemplar]" | Voice matching |
| "Push this further" | Intensity increase |
| "What am I missing?" | Gap identification |

### Prompt Debugging
| Problem | Solution |
|---------|----------|
| Too generic | Add specific constraints |
| Off-topic | Strengthen context |
| Wrong tone | Add voice examples |
| Too long | Add length constraint |
| Too short | Ask for expansion |
| Missing something | Provide exemplar |

---

*"The prompt is not a command. It is an invitation to co-create."*

Overview

This skill teaches the Arcanean Prompt Language (APL), an advanced prompt-engineering system that uses mythic frameworks, constraint architecture, and the Centaur Principle for collaborative human-AI creation. It provides templates, voice patterns, iteration protocols, and techniques to produce higher-quality outputs faster. The goal is practical: tighten prompts, amplify creative tension, and build reliable multi-step workflows for agents and creators.

How this skill works

APL structures prompts around five pillars: Identity, Context, Constraint, Exemplar, and Iteration. You define a speaking role, frame the setting and stakes, apply hard/soft/creative constraints, supply one excellent example, and run focused refinement loops. The skill also offers meta-prompting, persona stacking, temporal framing, and escalating constraints to iteratively improve results.

When to use it

  • Designing prompts for multi-agent orchestration or creative workflows
  • Improving AI output quality when results are vague or unfocused
  • Rapid prototyping of creative content with clear style and constraints
  • Teaching teams how to co-create with AI using repeatable protocols
  • Debugging prompts that produce hallucinations or off-tone responses

Best practices

  • Start by defining a precise Voice Identity with sample phrases and never-say items
  • Balance constraints: aim for ~60–70% defined, 30–40% exploratory
  • Always include one high-quality exemplar and one anti-exemplar when possible
  • Use short iteration cycles: generate, evaluate, tighten constraints, and regenerate
  • Prefer few-shot examples over long textual instructions for style learning

Example use cases

  • Create a branded content brief: identity=brand voice, constraints=format and length, exemplar=best-performing post
  • Run a multi-agent world-building session: persona-stack agents for lore, psychology, and narrative
  • Refine product copy via iteration commands like “Push this 20% further” and “Keep X, change Y”
  • Design technical prompts with hard constraints for format and API specs to avoid failure modes
  • Generate contrastive training examples (good vs. bad) to teach style and guardrails

FAQ

How many examples do I need to teach a style?

Three to five few-shot examples are usually enough for consistent style transfer; one excellent example plus one anti-exemplar works well for focused refinement.

When should I use hard vs soft constraints?

Use hard constraints for non-negotiables (format, length, legal limits). Use soft constraints for preferences like tone or pacing that can be nudged during iteration.