MIMIC-Py: An Extensible Tool for Personality-Driven Automated Game Testing with Large Language Models

Yifei Chen, Sarra Habchi, Lili Wei
Accepted by The 33rd ACM International Conference on the Foundations of Software Engineering
FSE Companion '26
Paper Demo Video Code arXiv MIMIC: The Research Paper

Dungeon Adventures Demo

Minecraft Demo

Shattered Pixel Dungeon Demo

Abstract

Modern video games are complex, non-deterministic systems that are difficult to test automatically at scale. Although prior work shows that personality-driven Large Language Model (LLM) agents can improve behavioural diversity and test coverage, existing tools largely remain research prototypes and lack cross-game reusability.

This tool paper presents MIMIC-Py, a Python-based automated game-testing tool that transforms personality-driven LLM agents into a reusable and extensible framework. MIMIC-Py exposes personality traits as configurable inputs and adopts a modular architecture that decouples planning, execution, and memory from game-specific logic. It supports multiple interaction mechanisms, enabling agents to interact with games via exposed APIs or synthesized code. We describe the design of MIMIC-Py and show how it enables deployment to new game environments with minimal engineering effort, bridging the gap between research prototypes and practical automated game testing.

MIMIC-Py

MIMIC Overview

MIMIC-Py consists of four components: the Planner, Action Summarizer, Action Executor, and Memory System.

The Planner is the core module, generating action plans from predefined personality traits and past experiences. These experiences are stored in the Memory System, with the Action Summarizer analyzing execution results to produce summaries. The Action Executor then translates the Planner’s output into in-game actions. In each action iteration, the Planner produces a plan, the Executor executes it, and the Summarizer records feedback as new memory to guide future planning. The following sections detail each component.

Personality Trait Definitions from PathOS

Personality Trait Definitions

Personality Trait Description
Curiosity Drive to explore the extents of the game world; particularly drawn to POIs and NPCs
Achievement Desire to earn game achievements and accomplish feats of skill, endurance, or luck; drawn to goals and collectibles
Completion Motivation to finish a game "100%", completing every checkbox along the way; drawn to most interactive objects
Aggression Wanting to exert dominance over the game world and seek out combat; particularly drawn to enemies
Adrenaline Seeking the thrill of challenge, looking to take risks; particularly drawn to environmental hazards and enemies
Caution Averse to danger; repelled by hazards, drawn to survival resources
Efficiency Desire to meet end objectives as quickly as possible; particularly drawn to mandatory and final goals
Samantha Stahlke, Atiya Nova, and Pejman Mirza-Babaei. 2020. Artificial Players in the Design Process: Developing an Automated Testing Tool for Game Level and World Design. In Proceedings of the Annual Symposium on Computer-Human Inter- action in Play (Virtual Event, Canada) (CHI PLAY ’20). Association for Computing Machinery, New York, NY, USA, 267–280. doi:10.1145/3410404.3414249

In-Game Elements / Entity Types

Element Name Description Refer to...
Optional Goal Objective marker not necessary for level completion (e.g., sidequest mission markers, optional puzzles) Objective, optional goal, goal point
Mandatory Goal Objective marker necessary for level completion (e.g., main mission marker) Objective, committed goal, goal point
Final Goal Objective which triggers the end of a given level Objective, outcome
Enemy Hazard A hostile character, etc. which could incite combat Enemies, agents, combat
Environment Hazard Physical hazard which could damage the player Deadly traps, obstacles
Survival Resource Item which can be picked up for some benefit to player survival, such as health packs Resource, power-up, resource
Collectible Item which can be picked up to contribute to achievements or unlocking content Resource, objective, collecting, pick-ups
Point-of-Interest (POI) Environmental feature, landmark, or setpiece intended to capture players’ visual interest Exploration
Non-Player Character (NPC) Non-hostile character who may be interacted with for story purposes, completing a mission, etc. Characters, agents
Samantha Stahlke, Atiya Nova, and Pejman Mirza-Babaei. 2020. Artificial Players in the Design Process: Developing an Automated Testing Tool for Game Level and World Design. In Proceedings of the Annual Symposium on Computer-Human Inter- action in Play (Virtual Event, Canada) (CHI PLAY ’20). Association for Computing Machinery, New York, NY, USA, 267–280. doi:10.1145/3410404.3414249

BibTeX

To be published in the Proceedings of the 34th ACM International Conference on the Foundations of Software Engineering, FSE Companion '26

@inproceedings{MIMIC_Py_YIFEI_FSE_26,
    title        = {MIMIC-Py: An Extensible Tool for Personality-Driven Automated Game Testing with Large Language Models},
    author       = {Chen, Yifei and Habchi, Sarra and Wei, Lili},
    year         = 2026,
    month        = {},
    booktitle    = {Proceedings of the 34th ACM International Conference on the Foundations of Software Engineering (FSE Companion '26)},
    volume       = {5},
    number       = {},
    pages        = {},
    doi          = {arXiv:2604.07752},
    issn         = {},
    abstract     = {Modern video games are complex, non-deterministic systems that are difficult to test automatically at scale. Although prior work shows that personality-driven Large Language Model (LLM) agents can improve behavioural diversity and test coverage, existing tools largely remain research prototypes and lack cross-game reusability. This tool paper presents MIMIC-Py, a Python-based automated game-testing tool that transforms personality-driven LLM agents into a reusable and extensible framework. MIMIC-Py exposes personality traits as configurable inputs and adopts a modular architecture that decouples planning, execution, and memory from game-specific logic. It supports multiple interaction mechanisms, enabling agents to interact with games via exposed APIs or synthesized code. We describe the design of MIMIC-Py and show how it enables deployment to new game environments with minimal engineering effort, bridging the gap between research prototypes and practical automated game testing. The source code and a demo video are available on our project webpage: https://mimic-persona.github.io/MIMIC-Py-Home-Page/},
    keywords     = {Artificial Intelligence; Human-Like Gaming Agents; PersonalityDriven Gaming Agents; Automated Game Testing; Large Language Models (LLMs)}
}