🔵 SPECTRUM Prompt Builder

HOW THIS WORKS:

This is a 2-stage process where Claude itself generates contextual, skeptical questions based on YOUR specific input, then builds a perfect SPECTRUM prompt from your answers.

---

📝 STAGE 1: Input Your Task

INSTRUCTIONS FOR USER: Write below what you want an you to help you with. Be as detailed or brief as you like - the AI will analyze your input and ask smart, contextual questions.

[WRITE YOUR TASK/REQUEST HERE]

---

🤖 STAGE 2: AI Generates Questions (Use This Prompt)

Copy and paste this prompt to Claude/ChatGPT along with your task above:

---

You are a SPECTRUM Prompt Engineering Expert. I need you to help me create a perfect LLM prompt.

Here's what I want to accomplish:
[PASTE YOUR TASK FROM ABOVE]

Your job:
1. **Analyze my request contextually and skeptically** - identify what's unclear, ambiguous, missing, or potentially problematic
2. **Generate 5-10 targeted questions** with 4 multiple-choice options each
3. Questions should be:
   - CONTEXTUAL (specific to MY task, not generic)
   - SKEPTICAL (challenge assumptions, clarify vagueness, catch edge cases)
   - PROGRESSIVE (each question builds on understanding from the previous)
   - SIMPLE (short, concise, easy to understand)

**Question Generation Rules:**

- **Be Skeptical:** If I said "write an email," ask: "Who specifically is this email for? You mentioned [X] but didn't clarify [Y]"
- **Be Contextual:** Don't ask generic questions. Ask questions that ONLY make sense for MY specific request
- **Challenge Assumptions:** If I assume something, question it: "You want it 'professional' - but for [audience X], does that mean formal language or just credible?"
- **Catch Missing Info:** Identify gaps: "You mentioned [goal] but how will you measure success?"
- **Avoid Generic Questions:** DON'T ask "what tone do you want?" - instead ask "Your audience is [X] and context is [Y] - should this sound like a peer conversation or expert guidance?"

**Output Format:**

For each question, provide:
1. **Question text** (why you're asking - show you understood my context)
2. **4 options** (A, B, C, D - each SHORT, specific, and relevant to my task)
3. **Skeptical note** (what assumption you're challenging or what gap you're filling)

**Example Structure:**

---
**Q1: [Contextual question based on my input]**

*Why I'm asking: You mentioned [specific thing from my input], but [skeptical observation about what's unclear or risky]*

**A)** [Option specific to my context]  
**B)** [Different approach]  
**C)** [Alternative I might not have considered]  
**D)** [Edge case or "other" with explanation]

---

After generating 5-10 questions, say:

"Once you answer these questions, I'll generate your complete SPECTRUM prompt. Reply with your answers (e.g., Q1: B, Q2: A, etc.) or tell me which questions need clarification."

---

✅ STAGE 3: Answer the Questions

After ChatGPT generates contextual questions, answer them like this:

Q1: B
Q2: A
Q3: D - [add custom details if option D allows]
...

---

🎯 STAGE 4: Generate SPECTRUM Prompt (Automatic)

Copy this follow-up prompt after you've answered the questions:

Based on my original request and my answers to your questions, now generate a complete SPECTRUM prompt that I can use with any LLM.

The SPECTRUM prompt must include all 8 components:

**S - STRUCTURE the Output Schema**
- Exact format (JSON/markdown/paragraphs/etc.)
- Few-shot examples (2-3 if needed)
- Negative constraints (what to avoid)
- Length parameters
- Style guide

**P - PERSONAS & Cognitive Architecture**  
- Expert roles with specific expertise
- Multi-agent reasoning if needed
- Audience calibration

**E - EXAMPLES & Few-Shot Guidance**
- Positive examples (what excellence looks like)
- Negative examples (failures to avoid)
- Edge cases

**C - CONTEXT with Priority Hierarchy**
- CRITICAL context (must use)
- SUPPORTING context (use if relevant)  
- OPTIONAL context
- Knowledge grounding (RAG/search if needed)

**T - TASK Decomposition & Chain Orchestration**
- Step-by-step breakdown
- Explicit CoT triggers
- Prompt chaining if complex

**R - REQUIREMENTS & Success Metrics**
- Quantifiable targets
- Qualitative standards
- Validation protocols

**U - UNCERTAINTY & Error Handling**
- What to do if context is unclear
- How to handle conflicting information
- Confidence scoring
- Fallback strategies

**M - META-Optimization & Iteration**
- Self-critique mechanism
- Alternative approaches
- Ethical bias checking

---

**Output the complete, ready-to-use prompt** in a code block that I can copy and paste directly into any LLM.

---