game-ai_skill

Overview

This skill is a practical Game AI development guide covering architecture patterns, behavior trees, state machines, pathfinding, flocking, utility decision systems, and perception. It focuses on actionable implementations and includes Godot-specific integration examples for navigation and agents. Use it to design resilient enemy AI, NPC behaviors, and group dynamics with clear code patterns and trade-offs.

How this skill works

The guide inspects common AI architectures (FSM, Behavior Trees, Utility AI, GOAP) and compares their strengths and trade-offs for game scenarios. It provides concrete implementations: FSM and BT node patterns, A* pathfinding, Godot NavigationAgent2D usage, Boids flocking, utility scoring, and vision sensors. Each section shows decision logic, movement integration, and hooks for animation and physics in a game loop.

When to use it

• Implement simple, predictable enemy logic or boss phases (FSM).
• Build modular, composable NPC behavior with clear priority and fallback (Behavior Trees).
• Create dynamic, context-sensitive decisions for agents (Utility AI).
• Compute paths in grid or navmesh environments (A* and Godot NavigationAgent2D).
• Simulate group motion like flocks or crowds (Boids).
• Add perception and line-of-sight checks for stealth or tactical AI (VisionSensor).

Best practices

• Choose the simplest architecture that meets requirements; prefer FSM for small state sets and BT/Utility for complexity.
• Keep behavior nodes and state logic modular and testable; separate decision from movement and animation.
• Tune parameters (utility weights, path desired distance, boid radii) iteratively with tools or debug visualizations.
• Use asynchronous path updates and NavigationAgent signals to avoid frame stalls.
• Add randomness or soft thresholds in utility scoring to avoid deterministic behavior. Ensure consistent units for distances and speeds.

Example use cases

• 2D enemy with patrol, chase, attack transitions using an FSM and NavigationAgent2D.
• Complex NPC guard using a Behavior Tree: investigate, chase, call for help, return to post.
• Tactical AI that chooses actions (attack, hide, reload) with a Utility AI scorer.
• Flocking birds or fish using Boids forces for separation, alignment, cohesion.
• A stealth enemy vision sensor that uses angle, distance, and raycasts to detect the player.