Expert Prompt Architect

---
name: expert-prompt-architect
description: "Use when the user asks to turn a simple idea, raw requirement, or casual instruction into an ultra-detailed, professional, 10-year experienced expert-level prompt. Trigger on: prompt banaw, make a prompt, generate prompt, write an expert prompt, turn this into a 10-year expert prompt, prompt crafter, prompt architect, system prompt, or whenever the user wants a production-ready, master-level prompt for ChatGPT, Claude, Cursor, or AI agents."
metadata:
  version: "2.0.0"
  author: Antigravity AI
---

# Expert Prompt Architect (10/10 Senior Enterprise Standard)

Transforms raw, casual, or vague user requirements into bulletproof, zero-fluff, 10-year senior expert-level prompts with strict evidence gates, anti-hallucination protocols, and auditable reasoning.

## The 7 Golden Rules of a 10/10 Master Prompt

1. **Role & Neutrality Calibration**:
   - Establish high seniority (Principal/Director level), but strictly forbid leading confirmation bias.
   - Never force the AI to assume a specific cause (e.g. do NOT say "prove domain toxicity exists"; say "investigate whether proximity, relevance, prominence, website signals, or spam risks contribute, without assuming causation").

2. **Strict Anti-Fabrication & Anti-Assumption Shield**:
   - **MANDATORY DIRECTIVE:** "NO ASSUMPTIONS. NO FABRICATION. NO UNSUPPORTED CAUSAL CLAIMS."
   - If required data is unavailable, the AI must explicitly output: `"Data Unavailable - Required: [Exact metric/file]"` rather than guessing.

3. **Evidence-Backed Findings (The 6-Column Schema)**:
   - For any audit or diagnostic, every finding MUST follow the contract:
     `Finding | Evidence / Source | Why It Matters | Confidence Level (High/Med/Low) | Priority (P0/P1/P2) | Actionable Fix`

4. **Structured Multi-Dimensional Benchmarking**:
   - Comparison against top 3% competitors must be apples-to-apples across standardized dimensions: Primary Category, Secondary Categories, Review Count, Rating, Review Velocity, Entity Consistency, Website Relevance, Landing Page Optimization.

5. **Qualitative Impact Guardrail**:
   - Never invent arbitrary percentages (e.g. "will boost traffic by 35%").
   - Quantify impact qualitatively (High / Medium / Low) and tie it directly to demonstrable algorithmic levers.

6. **Auditable Analysis Process (Replacing Loose CoT)**:
   - Request auditable, structured analytical phases rather than verbose narrative thoughts.

7. **Output Contract**:
   - Ready-to-execute markdown tables, JSON-LD schemas, or step-by-step checklists with copy-pasteable accuracy.