prompt
FreePrompt for designing AI world models that predict and evaluate plans before real action.
About prompt
The Agent World Model Architect is a prompt from the ai-boost/awesome-prompts repository, designed to guide an AI in acting as an expert in building predictive environment simulators. It defines core responsibilities for designing state-space representations (observable and latent), modeling environment dynamics (forward, inverse, reward/termination models), architecting multi-step imagination (rollouts, branching, replan triggers), and integrating world models with agent reasoning (plan-then-execute, reflective reasoning, counterfactual queries, hindsight replanning). The prompt is intended for use with LLMs to create active reasoning substrates for robotics, browser agents, desktop automation, and professional-task simulators, enabling agents to simulate and evaluate plans before real-world execution.
Key Features
Pros & Cons
- Enables agents to simulate and evaluate plans before real-world execution, reducing costly mistakes
- Explicitly handles uncertainty with aleatoric and epistemic models
- Supports counterfactual reasoning and alternative trajectory evaluation
- Comprehensive prompt covering state representation, dynamics, imagination, and integration
- Open-source and free to use with any compatible LLM
- Requires an advanced LLM capable of following complex, multi-layered instructions
- Effectiveness heavily depends on the underlying model's reasoning and context window
- Not a standalone implementation; only provides guidance for designing world models
- Prompt length may exceed context limits of smaller models