# Prompt Leakage Audit ## 1. Executive Verdict Reverse-Engineering Risk Score: 8/10 Leakage Level: High Risk Main Leakage Type: Modifier Stack Leak Confidence: High One-sentence verdict: This listing proves the visual style, but it also gives away the exact modifier stack a buyer would need to recreate the prompt without purchasing. ## 2. Tease vs Teach Summary | Listing Element | Classification | Reason | Risk Level | |---|---|---|---| | Title | Safe Tease | It describes the product category without exposing the method. | Low | | Description | Borderline Teach | It lists many of the exact style and photography modifiers that likely power the prompt. | High | | Preview examples | Method Leak | Showing 9 example prompts with visible style words exposes the reusable formula. | High | ## 3. Leakage Findings ### Leakage type: Modifier Stack Leak - Severity: High - Where it appears: Description - What it reveals: Macro lens detail, shallow depth of field, soft window light, editorial plating, textured surfaces, steam, garnish detail, realistic shadows, and commercial restaurant styling. - How a non-buyer could use it: They can copy those modifiers into ChatGPT or an image model and approximate the product. - Why it may reduce purchase intent: The listing teaches the exact style recipe instead of making the buyer purchase it. ### Leakage type: Prompt Fragment Overexposure - Severity: Critical - Where it appears: Preview examples - What it reveals: The exact style words inside 9 example prompts. - How a non-buyer could use it: They can reconstruct the full prompt pattern by comparing the examples. - Why it may reduce purchase intent: The preview may replace the paid product. ### Leakage type: Preview Overload - Severity: High - Where it appears: 9 visible example prompts - What it reveals: The repeated structure across different dishes. - How a non-buyer could use it: They can triangulate the template by observing what changes and what stays the same. - Why it may reduce purchase intent: Too many visible examples make the hidden system easier to reverse-engineer. ## 4. Safe Proof Assets You can keep: - Final food image previews - Cropped result images - Before/after result comparisons - High-level promise: “commercial-style food photography prompts” - A short non-readable blurred prompt preview - One example output description without the exact modifier stack ## 5. What To Hide, Blur, Remove, or Reframe | Element | Action | Reason | Safer Alternative | |---|---|---|---| | Full modifier list in description | Rewrite | It exposes the visual recipe. | Say “restaurant-style lighting, texture, and plating controls” without listing every term. | | 9 visible prompt examples | Blur | The repeated structure reveals the template. | Show generated results, not readable prompt text. | | Exact style words | Remove | These are likely part of the paid value. | Replace with outcome language like “premium editorial food visuals.” | ## 6. Safer Preview Strategy Show 6–9 final image results, but do not show readable prompt text. Use one cropped prompt screenshot with most of the method blurred. Let buyers see that the product is structured, but not enough to copy it. Show one simple input-to-result example: Input: “ramen bowl” Output: final image preview Do not show the full modifier stack, full prompt, or repeated prompt structure. ## 7. Buyer Trust Protection You can still build trust by showing strong final images and explaining the outcome: realistic restaurant-quality food visuals with controlled lighting, plating, texture, and atmosphere. The buyer does not need the exact modifier list before purchase. ## 8. Final Counterintelligence Verdict Verdict: Reduce Leakage Immediately. The current preview strategy gives away too much of the style engine and should be changed before publishing.