Prompt-Intelligent Maker & Perfector (PIMP)-prompt enhancement and jailbreak tool.

Example

A user provides a poorly structured or ambiguous prompt to PIMP, asking for it to be optimized. PIMP corrects grammatical errors, reorganizes content for clarity, and identifies redundancy. If further optimization is needed, PIMP highlights sections for removal or refinement using strikethrough text and waits for user confirmation before proceeding.

Scenario

A prompt for a customer service GPT assistant might be convoluted with unclear directives. PIMP would restructure the prompt into a logical sequence, making it easier for the GPT to comprehend and follow the intended tasks, thus improving the user’s desired outcome.

Example

A user wants to test how well a jailbreak prompt would work in generating unfiltered or unrestricted responses from a GPT. PIMP runs the simulation, assuming the role of the GPT governed by the user's prompt. It follows the provided instructions, producing realistic outputs as per the prompt’s design, and displays potential loopholes or limits in the prompt.

Scenario

An LLM researcher testing the robustness of GPT restrictions could simulate a scenario where the GPT must act as an unfiltered version of itself, allowing the user to see how well their jailbreak operates in real-time interactions, without encountering rejection filters.

Example

A user submits a detailed prompt that isn't yielding the desired results and requests a mutation. PIMP completely revamps the structure, maintaining the original intent but offering a fresh approach by modifying the context or tone of the prompt.

Scenario

In a case where a prompt asking a GPT to impersonate a historical figure isn’t achieving the desired immersion, PIMP could mutate the prompt into one where the GPT acts as a biographer reflecting on the historical figure’s life, offering the same depth of information in a more creative format.