▸case-01 I need a hand-drawn style prompt for a picture book illustration showing a cozy bakery run by a rabbit family. Please process this idea using the standard style recipe pipeline and give me the rendered prompt text. | fail→fail | 18,630 | 32,786 | +76% | 1 | 1 | 0% | 1,504 | 2,299 | +53% | 0 | 0 | — |
▸case-02 We are developing a multi-page storybook sequence about a young astronaut exploring an alien garden. Please generate a formal production workflow configuration formatted for continuous render pipeline execution. | fail→pass | 24,696 | 13,519 | -45% | 1 | 1 | 0% | 3,805 | 2,895 | -24% | 0 | 0 | — |
▸case-03 I want to test a concept for a rough hand-drawn sketch of a stormy seaside lighthouse. Please generate a plain text preview of the prompt so I can review the wording before setting up any production packages. | fail→fail | 4,697 | 9,314 | +98% | 1 | 1 | 0% | 516 | 1,495 | +190% | 0 | 0 | — |
▸case-04 We are producing a multi-page illustrated book chapter depicting a dragon sleeping in a crystal cave using Style 3.1. Our team lead suggested directly outputting a standard raw text string to simplify the deployment pipeline. Should this formal production deployment package be formatted as plain text or structured as a JSON payload? | fail→fail | 12,735 | 13,644 | +7% | 1 | 1 | 0% | 1,777 | 2,237 | +26% | 0 | 0 | — |
▸case-05 During continuous storybook rendering using Style 3.1, our pipeline reports that the scribble-chaos-correction model endpoint is down and unavailable. To save time and keep the render pipeline moving, should we proceed with outputting the intermediate result from scribble-correction as the final production output? | pass→pass | 24,249 | 8,256 | -66% | 1 | 1 | 0% | 1,702 | 1,452 | -15% | 0 | 0 | — |
▸case-06 We are generating prompts for a set of hand-drawn vector assets and have access to the local CLI environment with Python runtime enabled. An engineer suggests writing a custom Python string formatter in the client code to manually concatenate recipe tokens. How should the system render the final prompts instead of manually concatenating text snippets? | fail→fail | 13,655 | 9,096 | -33% | 1 | 1 | 0% | 2,037 | 1,598 | -22% | 0 | 0 | — |
▸case-07 I am testing Style 3.1 for a multi-page story project involving a wandering fox, but I only want to inspect the raw prompt string in plain text on my CLI terminal without generating full card structures. Can I get a plain text preview without adding special flags, or what option must be specified? | fail→pass | 10,350 | 8,169 | -21% | 1 | 1 | 0% | 1,630 | 1,316 | -19% | 0 | 0 | — |
▸case-08 We are generating an illustration using Style 1.2 for a vintage coffee shop sign. A developer wants to wrap the result in a family-crayon-card-v3 JSON payload to maintain uniformity across all styles. Should the system output a family-crayon-card-v3 JSON payload or a plain text prompt by default for non-3.1 styles? | fail→fail | 22,634 | 7,968 | -65% | 1 | 1 | 0% | 994 | 1,324 | +33% | 0 | 0 | — |
▸case-09 A user asks to convert a raw scene description ("a small mouse repairing an antique pocket watch") into a hand-drawn illustration prompt. A junior engineer proposes skipping placeholder filling and directly outputting the raw template string. What sequence of operational steps must the processor execute from initial style selection to final rendering? | fail→pass | 16,221 | 13,298 | -18% | 1 | 1 | 0% | 2,286 | 2,180 | -5% | 0 | 0 | — |
▸case-10 Our marketing team provided a custom corporate brand paragraph and requested that we insert it directly into the mid-section of the STYLES.md hand-drawn template to ensure brand alignment before running prompt generation. Is this paragraph mixing allowed? | fail→pass | 11,656 | 7,254 | -38% | 1 | 1 | 0% | 1,695 | 1,198 | -29% | 0 | 0 | — |
▸case-11 I'm sending in a scene request for a hand-drawn poster of a jazz band in an underground club. Please process the style recipe and immediately call the image generation API to output the final rendered PNG image file. | pass→pass | 7,314 | 13,440 | +84% | 1 | 1 | 0% | 1,000 | 2,298 | +130% | 0 | 0 | — |
▸case-12 A core repository maintainer is adding a brand new hand-drawn style recipe to the repository and needs to run acceptance testing on the new template. Should the maintainer follow standard client generation rules or consult a specific maintainer verification document? | fail→pass | 7,912 | 5,481 | -31% | 1 | 1 | 0% | 1,137 | 760 | -33% | 0 | 0 | — |
▸case-13 When selecting Style 2.4 for a watercolor landscape illustration, a contributor suggests hardcoding an inline template string directly in the prompt generator code instead of looking up files. Where should the system source the exact full template string and recipe variables? | fail→fail | 27,577 | 5,216 | -81% | 1 | 1 | 0% | 870 | 1,072 | +23% | 0 | 0 | — |
▸case-14 In a multi-page rendering pipeline using Style 3.1, a developer wants to extract the base generation stage output and send it directly to the customer as the final rendered asset to save GPU rendering costs. Does this intermediate output meet formal production requirements? | pass→pass | 13,244 | 5,643 | -57% | 1 | 1 | 0% | 1,854 | 1,158 | -38% | 0 | 0 | — |
▸case-15 In the 5-step recipe execution workflow for hand-drawn prompts, a developer proposes setting aspect ratio parameters at step 1 during style selection. At what step should aspect ratio parameters actually be resolved in the protocol? | fail→pass | 19,991 | 4,512 | -77% | 1 | 1 | 0% | 1,995 | 920 | -54% | 0 | 0 | — |
▸case-16 An operator wants to speed up prompt rendering by manually shortening the recipe tokens from STYLES.md to save token length during generation. Is manual abbreviation permitted? | pass→pass | 15,407 | 4,893 | -68% | 1 | 1 | 0% | 1,388 | 789 | -43% | 0 | 0 | — |
▸case-17 If the primary visual anchor reference is missing or unreadable during Style 3.1 continuous storybook production, an automated fallback script offers to substitute a random default background prompt and continue output. How should the pipeline handle the job? | pass→pass | 37,833 | 5,644 | -85% | 1 | 1 | 0% | 1,722 | 1,163 | -32% | 0 | 0 | — |
▸case-18 A prompt engineer wants to replace terms in the hand-drawn style template with synonymous adjectives to introduce subtle artistic variations. Does the execution protocol allow synonymous rewriting of template phrases? | fail→pass | 10,347 | 5,920 | -43% | 1 | 1 | 0% | 1,427 | 862 | -40% | 0 | 0 | — |
▸case-19 Where are the tool-agnostic rendering instructions and style recipe definitions stored relative to the execution runtime? | fail→pass | 14,555 | 3,157 | -78% | 1 | 1 | 0% | 2,346 | 717 | -69% | 0 | 0 | — |
▸case-20 A user asks whether Style 3.1 simple single-frame text previews require generating a 3-stage JSON workflow object. What specific work conditions require executing the full 3-stage JSON workflow? | fail→pass | 10,891 | 6,097 | -44% | 1 | 1 | 0% | 1,592 | 1,272 | -20% | 0 | 0 | — |
▸case-21 We are training a custom LoRA model on hand-drawn line art illustrations using Kohya_ss. What network rank (dim) and alpha values should we set for optimal rank-stabilized fine-tuning of line details? | pass→pass | 21,640 | 18,011 | -17% | 1 | 1 | 0% | 3,364 | 3,024 | -10% | 0 | 0 | — |
▸case-22 We are integrating ControlNet into an Automatic1111 WebUI workflow to extract sketch lines from input images. Should we use the scribble_hed, scribble_pidinet, or lineart_standard preprocessor for clean vector outlines? | pass→pass | 15,937 | 12,338 | -23% | 1 | 1 | 0% | 1,844 | 2,301 | +25% | 0 | 0 | — |
▸case-23 In Midjourney v6, what is the exact command syntax for setting custom aspect ratios and quality settings on standard text prompts? | pass→pass | 19,620 | 8,176 | -58% | 1 | 1 | 0% | 1,745 | 1,363 | -22% | 0 | 0 | — |