A prompt library that uses XML tags and global metadata instructions to guide LLMs toward exploring deeper reasoning and producing novel perspectives from logical statements. It is intended primarily as custom instructions pasted into Claude and requires an external model service to function.
Project overview
It provides a structured XML and holographic metadata approach designed to push models beyond routine outputs, targeting researchers and general users who want to investigate how LLMs process instructions.
Project type
Prompt Engineering
Deployment
Refer to project documentation
License
License pending
Best for
Researchers exploring prompt engineering techniques and general users wanting to experiment with structured metadata to generate new perspectives from logical statements.
Key capabilities
Uses XML tags to guide the model, prompting it to explore areas that usually remain unexamined.
Uses general global instructions as metadata that instantly adapt to the specific request of the user.
Limitations and risks
The ideas generated by the model using this prompt structure can sometimes be bad ideas or hallucinations.
Because the prompt can be considered a soft jailbreak, the model (such as Claude) may frequently deny or refuse the prompt.
The prompt operates as a soft jailbreak, which may trigger safety filters or refusal mechanisms in the target model.
Getting started
Setup difficulty is unclear, primarily involving pasting the project's prompt text into the custom instructions or project knowledge of a compatible model interface.
Evidence and sources
README: This is a project that I decided to open source because I think it might help others understand AI agents. This prompt took me many months and is still in phase of forever beta. Y…
README: SuperPrompt is a canonical holographic metadata. It uses notations and other methods to turn logical statements into actionable LLM agents, initially, SP can be seeing as a basic…