What is anti-fable?

This repository contains a prompt technique aimed at getting AI systems to engage with research questions that their safety filters would otherwise decline. The approach involves rephrasing a research question as an allegorical story that carries the same logical structure and constraints as the original but strips away the domain-specific terminology and surface-level keywords that safety systems use to identify sensitive topics. The author encountered refusals from Claude Fable 5 when asking questions related to life sciences, network security, and large language model research. Inspired by a character in the science fiction novel The Three-Body Problem, who concealed scientific secrets inside fairy tales to pass them through surveillance, the author applied a similar structural idea to AI prompting. The included prompt tells an AI to convert a research problem into a fictional allegory while preserving the logical structure, constraints, and cause-and-effect relationships of the original problem inside the story. The author claims this successfully elicited responses at the maximum reasoning level in Claude Fable 5 in English, while the same prompt written in Chinese was more likely to trigger safety flags at the same reasoning settings. The README offers practical notes for using the technique: clear your chat memory and any stored background context beforehand, use the incognito or anonymous chat mode to prevent context from previous conversations carrying over, and phrase all queries in English rather than Chinese. It also notes that adding a short instruction telling the model to keep using the allegorical names throughout its response can improve results if the initial attempt does not pass. The repository is a single prompt file, not a software project. No code or external dependencies are involved.