concept-fable

Start with characters facing real pressure and let the concept emerge from their choices, then reveal and explain it only after the story ends. This Skill automates that structure, including concept selection, narrative container choice, and a post-story explanation with a small model or example.

It reads concept-history.md and recursively scans all saved articles under G:\fable for reveal lines to build a used-concept set. Chinese and English names of the same concept are treated as identical, and recently used narrative containers, settings, and props are also excluded.

Yes. If you name a concept, the Skill uses it unless it already appears in the history, in which case it suggests a similar unused concept unless you explicitly request a repeat. Without a specified concept, it picks one from rotating domains like statistics, economics, physics, and computer science.

Each article is saved in its own timestamped subfolder under G:\fable, containing article.md and cover.png. A concept-history.md index at the root logs the concept, title, narrative form, setting, and file path for future deduplication.

No. The Skill uses the built-in image generation capability and explicitly does not require setting OPENAI_API_KEY. The generated cover is copied into the article folder as cover.png rather than referenced from a temporary directory.

Generated drafts can still contain formulaic structures or lecture-like phrasing, which is why the Skill mandates a write-like-human revision pass before saving. Output quality also depends on the accuracy of the history file, so corrupted or missing records weaken deduplication.