mcpbeat Sign in

Bevy Character Movement Agent Skill

Bevy 0.20 kinematic character movement: intent components, FixedUpdate integration, coyote time, and jump buffer. Use when a player or actor walks or jumps and the step must stay stable. Keywords: Bevy, movement, FixedUpdate, coyote, jump buffer, ECS.

2.5k tokens
context cost
SKILL.md, read when the skill is used
2
files
plus reference files
0
copies elsewhere
how many repositories repackaged it
854
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/thedivergentai/Game-Design-Agentic-Skills --skill bevy-character-movement

What comes with it

537 bytes besides the instruction
references/ecs-ownership.md

The instruction itself

16 sections, as written by the author

Bevy character movement

When to use

Load this when a Bevy app needs a player or actor that moves in discrete steps (walk, jump feel, ground check) and you need the ECS shape before any solver code exists. Coyote time and a jump buffer are design notes. They are not a node port.

ECS shape

  • Components own pose and intent: position, velocity, and a jump-buffer or coyote timer as data. The Quick Start ECS page shows components as Rust structs that derive Component.
  • Resources own clocks and tuning shared by every mover. bevy::time 0.20.0 documents Time, Fixed (fixed timestep clock), and Virtual.
  • Systems read intent and write velocity or position. Register them on the fixed schedule, not on a per-node physics callback.
  • Schedule: run_fixed_main_schedule runs FixedMain from virtual time (bevy::time 0.20.0). Order dependent systems with .chain() (ECS page). Do not paste a Godot physics-process function.

Engine counterparts

| Godot idea | Bevy idiom |

| --- | --- |

| Body that you move yourself | Entity with movement components and a system |

| Step that should stay stable when frame time changes | System on the fixed timestep (Fixed / FixedMain) |

| Coyote time, jump buffer | Timer data on the entity or a small resource, consumed by the movement system |

| Collision response | Collision response belongs on bevy-physics-3d. |

Scope

Locomotion here is kinematic data on FixedUpdate: intent, velocity, coyote time, and a jump buffer. Collision response belongs on bevy-physics-3d. Do not call move_and_slide, _physics_process, or CharacterBody2D.

[dependencies]
bevy = "0.20"
#[derive(Component, Default)]
struct PlanarIntent {
    axis: Vec2,
    jump_pressed: bool,
}

#[derive(Component)]
struct KinematicBody {
    velocity: Vec3,
    grounded: bool,
    coyote: f32,
    jump_buffer: f32,
}

#[derive(Resource)]
struct MoveTuning {
    speed: f32,
    jump_speed: f32,
    gravity: f32,
    coyote_time: f32,
    jump_buffer_time: f32,
}

impl Default for MoveTuning {
    fn default() -> Self {
        Self {
            speed: 6.0,
            jump_speed: 8.0,
            gravity: 20.0,
            coyote_time: 0.12,
            jump_buffer_time: 0.12,
        }
    }
}

fn read_move_intent(keys: Res<ButtonInput<KeyCode>>, mut bodies: Query<&mut PlanarIntent>) {
    let mut axis = Vec2::ZERO;
    if keys.pressed(KeyCode::KeyA) {
        axis.x -= 1.0;
    }
    if keys.pressed(KeyCode::KeyD) {
        axis.x += 1.0;
    }
    let jump_pressed = keys.just_pressed(KeyCode::Space);
    for mut intent in &mut bodies {
        intent.axis = axis;
        intent.jump_pressed = jump_pressed;
    }
}

fn step_kinematic(
    time: Res<Time<Fixed>>,
    tuning: Res<MoveTuning>,
    mut bodies: Query<(&mut PlanarIntent, &mut KinematicBody, &mut Transform)>,
) {
    let dt = time.delta_secs();
    for (mut intent, mut body, mut transform) in &mut bodies {
        body.velocity.x = intent.axis.x * tuning.speed;
        if body.grounded {
            body.coyote = tuning.coyote_time;
        } else {
            body.coyote = (body.coyote - dt).max(0.0);
            body.velocity.y -= tuning.gravity * dt;
        }
        if intent.jump_pressed {
            body.jump_buffer = tuning.jump_buffer_time;
            intent.jump_pressed = false;
        } else {
            body.jump_buffer = (body.jump_buffer - dt).max(0.0);
        }
        if body.jump_buffer > 0.0 && body.coyote > 0.0 {
            body.velocity.y = tuning.jump_speed;
            body.jump_buffer = 0.0;
            body.coyote = 0.0;
            body.grounded = false;
        }
        transform.translation += body.velocity * dt;
        if transform.translation.y <= 0.0 {
            transform.translation.y = 0.0;
            body.velocity.y = 0.0;
            body.grounded = true;
        }
    }
}

fn spawn_mover(mut commands: Commands) {
    commands.spawn((
        Transform::default(),
        PlanarIntent::default(),
        KinematicBody {
            velocity: Vec3::ZERO,
            grounded: true,
            coyote: 0.0,
            jump_buffer: 0.0,
        },
    ));
}

pub struct CharacterMovementPlugin;

impl Plugin for CharacterMovementPlugin {
    fn build(&self, app: &mut App) {
        app.init_resource::<MoveTuning>()
            .add_systems(Startup, spawn_mover)
            .add_systems(Update, read_move_intent)
            .add_systems(FixedUpdate, step_kinematic);
    }
}

Sample ButtonInput::just_pressed in Update. FixedUpdate can run zero times or several times in one frame, so a press read there is dropped or repeated. The intent component keeps the edge until FixedUpdate consumes it.

When a physics skill writes Grounded, chain the coyote refill after that write so the timer sees the contact from the same tick.

#[derive(Component)]
struct Grounded;

#[derive(Component)]
struct CoyoteTimer(f32);

#[derive(Resource)]
struct CoyoteWindow(f32);

fn refill_coyote(window: Res<CoyoteWindow>, mut movers: Query<&mut CoyoteTimer, With<Grounded>>) {
    for mut timer in &mut movers {
        timer.0 = window.0;
    }
}

fn tick_coyote(time: Res<Time<Fixed>>, mut movers: Query<&mut CoyoteTimer, Without<Grounded>>) {
    let dt = time.delta_secs();
    for mut timer in &mut movers {
        timer.0 = (timer.0 - dt).max(0.0);
    }
}

pub struct CoyotePlugin;

impl Plugin for CoyotePlugin {
    fn build(&self, app: &mut App) {
        app.insert_resource(CoyoteWindow(0.12)).add_systems(
            FixedUpdate,
            (refill_coyote, tick_coyote).chain(),
        );
    }
}

Expert Knowledge Delta

Target: Bevy 0.20+.

Core ECS types: Component, Resource, Query, Commands, Startup, Update, Fixed / FixedMain (bevy::time), .chain(), Plugin, DefaultPlugins, MinimalPlugins. Messages and observers are an ownership slot. See bevy-events-messaging for their APIs. Systems run in parallel unless .chain() orders the ones that conflict.

Who owns state

| Piece | Owner | Who may mutate | Schedule |

| --- | --- | --- | --- |

| Data on one entity | Component | One system via Query | Update or FixedMain |

| One value for the whole app | Resource | One system | Startup to insert, then the schedule that reads it |

| Spawn | Commands | The spawning system | Startup for the initial world; later schedules for runtime spawn |

| Order that must not race | .chain() | The chained systems only | Same schedule |

| Plugin boundary | Plugin | build registers systems and resources | Before the app runs |

Borrow rule: two systems that both mutably query the same component must be ordered (.chain()) or split so only one writes. Unrelated systems stay parallel. That is the ECS page rule, not a guess at a system-set API.

Decision

| If you need | Do this | Do not |

| --- | --- | --- |

| Per-entity data | Component + Query | A Godot node script |

| App-wide config or clock | Resource | Copy the value onto every entity |

| Stable simulation step | FixedMain and the Fixed clock | Tie simulation to raw frame Update |

| Startup spawn | Commands in Startup | Spawn from a render system |

| Cross-system notice that is not shared state | Message or observer slot in bevy-events-messaging | A new event type invented in this file |

| An API outside this skill | Load the skill that owns it | A made-up crate or API |

Load references/ecs-ownership.md for the ownership table.

NEVER

  • Do not pin /0.20.0/ in docs URLs. WHY: links stay on latest; the 0.20 target is stated once in prose.
  • Do not step locomotion on raw Update when the step must stay stable. WHY: bevy::time documents Fixed and run_fixed_main_schedule (FixedMain) for a fixed clock.
  • Do not read ButtonInput::just_pressed inside FixedUpdate. WHY: that schedule can skip a frame or run more than once, so the press is lost or applied twice.
  • Do not call move_and_slide, _physics_process, or CharacterBody2D. WHY: those are Godot types. This skill only owns ECS movement data.
  • Do not invent a character-controller or shape-cast API. WHY: this skill owns kinematic intent and the fixed step; contact queries live on the physics skills.
  • Do not mutably query the same movement component from two unordered systems. WHY: systems run in parallel unless .chain() orders them.

Reference

The library pin on bevy-master stays 0.20.0. Official pages stay on the unversioned site.

Official Documentation

  • Quick Start: ECS - Entities, Component, systems, Query, Commands, and .chain().
  • Fixed timestep - Fixed and FixedMain via run_fixed_main_schedule on the ECS schedule page.
  • Examples - Example tree on the latest branch.
Prerequisites
  • bevy-master - ECS ownership, schedules, and which Bevy skill to load.
Complements
Downstream / consumers
  • Genre skills cover loop and encounter design.
Master

How to use it

Copy the folder

Take thedivergentai/bevy-character-movement from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.