What problem does it solve? Image generation models often garble text, invent numbers, and produce inconsistent results across related images. This Skill provides operational guidance for deciding when to generate images with AI tools versus building visuals with code, how to write effective prompts, and how to handle real people, factual grounding, and multi-image consistency. ## Core Features & Use Cases - Tool Selection Guidance: Clear rules for choosing between image_gen (new images from text) and image_edit (modifying existing images), including a decision table and rules of thumb. - Accuracy-First Workflows: Directs building charts, diagrams, and text-heavy visuals with HTML/CSS code instead of image models, with a verify-and-fix loop for discrete content. - Reference-First Real People Handling: Enforces web search verification and reference-image-based editing for named real people, with safety guardrails. - Video Shot Planning: Covers image_to_video workflows with shot-by-shot planning, 6s/10s durations, and FFmpeg stream-copy assembly. - Use Case: When asked to create a poster of a public figure, search the web to verify facts, use image_edit with a real reference photo, and verify the output text is accurate before delivering. ## Quick Start Load this skill whenever you are about to make an image_gen or image_edit call and follow its prompt-craft and tool-selection rules.
