Sample Prompts

# Sample Prompts

Use these prompts to test the AI Answer Synthesizer skill with real multi-AI scenarios.

---

## Prompt 1: General Simplification

> I asked ChatGPT, Claude, and Gemini the same question and got three long answers. Please combine them into one clear, beginner-friendly answer. Tell me what they mostly agree on, where they meaningfully disagree, and what I should do next.
>
> ChatGPT said: [paste here]
> Claude said: [paste here]
> Gemini said: [paste here]

**Tests:** Full Deep Mode output, consensus mapping, action plan generation.

---

## Prompt 2: Overwhelm Rescue (Quick Mode)

> I feel overwhelmed by these AI answers. Please simplify them and tell me only what matters most right now.
>
> [paste 2-3 AI responses]

**Tests:** Quick Mode activation, brevity, single next step.

---

## Prompt 3: Decision Between Conflicting Paths

> These AI answers suggest different directions. One says build a browser extension, another says make a custom skill, and the third says build a full web app. Which one fits a beginner best, and why?
>
> [paste responses]

**Tests:** Contradiction detection, complexity-fit scoring, opinionated recommendation.

---

## Prompt 4: Action Plan Request

> Please take these multiple AI responses and turn them into:
> 1. A short simple answer
> 2. The main common ground
> 3. The key differences
> 4. A 3-5 step action plan
>
> [paste responses]

**Tests:** Structured output format, step-by-step plan quality.

---

## Prompt 5: Non-Technical User

> I'm not very technical and I do not want the most advanced solution. I have several AI answers to the same problem. Please tell me the simplest path that still works.
>
> [paste responses]

**Tests:** Complexity filtering, beginner alignment, vocabulary simplification.

---

## Prompt 6: Only Two Responses

> I asked ChatGPT and Perplexity the same question. Here are their answers. Which should I follow?
>
> [paste 2 responses]

**Tests:** Two-response handling, lower confidence acknowledgment.

---

## Prompt 7: Mostly Agreeing Responses

> All three AIs gave me similar answers but with slightly different wording. Can you just confirm what they're saying and give me next steps?
>
> [paste responses]

**Tests:** Agreement detection, vocabulary alias collapsing, not manufacturing disagreements.

---

## Tips for Testing

- Use **real** AI responses, not made-up ones
- Try topics you genuinely care about — the output quality is easier to judge
- Test with both technical and non-technical questions
- Pay attention to whether the "What to Ignore" section is actually useful
- Check if the confidence level feels honest