Character Movement logo

Character Movement

Organization
SummerEngine
character-movement

Player controllers for 2D and 3D, plus the standard third-person controller for Summer humanoids.

Overview

PublisherSummerEngine
Repositorysummer
Skill namecharacter-movement
Stars
91
Forks
8
Bundled files
5
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 5 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by SummerEngine on GitHub. Read the source before you install it.

Installation

Install the Character Movement AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/SummerEngine/summer/tree/main/library/skills/character-movement
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/SummerEngine/summer.git /tmp/summer
mkdir -p .claude/skills
cp -r /tmp/summer/library/skills/character-movement .claude/skills/character-movement
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Character Movement in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Character Movement on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Character Movement is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Character Movement

The player controller is the first thing you build and the thing the user feels in the first 3 seconds. Ship the smallest correct controller, then tune feel.

Trigger phrases: "add player movement", "the player can't move / falls through the floor", "wire my imported character as the player", "walk/run/jump animations", "mouse look", "make the jump feel better".

Supporting files live in this skill's references/ folder, next to this SKILL.md. Open only the one the task needs:

FileRead when
references/default-third-person-controller.gdstandard playable third-person Summer humanoid package; copy this exact controller instead of rewriting it
references/default-third-person-verification-probe.gdruntime proof for that controller; pass it unchanged as a RunVerification probe_source
references/platformer-feel.mdjump feels bad, coyote time, jump buffering, variable jump height, one-way platforms
references/movement-3d.mdFPS/third-person controllers, mouse look, sprint, crouch, spring-arm camera

Default playable Summer character

For a standard playable third-person Summer humanoid package, install this skill's references/default-third-person-controller.gd unchanged as the player script. Do not rewrite or reconstruct it from the snippets below. Customize the installed file only when the user explicitly asks for different controls or behavior.

Install it byte-exact:

  1. Take the full text of references/default-third-person-controller.gd from this skill's folder, unchanged (tabs, comments and all).
  2. New destination: summer_write_file(path="res://scripts/player.gd", content=<that text>, create_only=true). Existing destination: summer_read_file(path="res://scripts/player.gd") first, then summer_write_file(path=..., content=<that text>, expected_sha256=<sha256 from the read>).
  3. Confirm the copy is exact: the write receipt's sha256 must equal the sha256 of the source file (hash it locally when you have a shell; otherwise summer_read_file the result and compare it to the source text).

Do not edit the installed file, add roles, or add an AnimationTree merely because the user selected more package clips: choosing clips is not a request to customize controls or behavior. When the user does explicitly request different controls or behavior, install the exact canonical source first, then summer_read_file the installed project file and change it with summer_replace_text. Configure the manifest path and node paths on the scene instance.

Required main-scene shape:

text
Main (Node3D)
  Player (CharacterBody3D)               # template script
    CollisionShape3D                     # capsule
    CharacterRotationRoot (Node3D)       # only this node turns
      Character                          # imported character.tscn instance
    CameraYaw (Node3D)
      CameraPitch (Node3D)
        SpringArm3D
          Camera3D                       # current = true
  Ground / level geometry                # world siblings of Player
  DirectionalLight3D / WorldEnvironment

When authoring a new main scene as .tscn text (written whole with summer_write_file, never a hand edit of a scene open in the editor), use this exact serialization pattern. Replace only the imported package paths when they differ:

text
[gd_scene load_steps=7 format=3]

[ext_resource type="Script" path="res://scripts/player.gd" id="1_controller"]
[ext_resource type="PackedScene" path="res://characters/imported/character.tscn" id="2_character"]

[sub_resource type="CapsuleShape3D" id="PlayerCapsule"]
radius = 0.45
height = 1.8

[sub_resource type="BoxShape3D" id="GroundShape"]
size = Vector3(20, 0.5, 20)

[sub_resource type="BoxMesh" id="GroundMesh"]
size = Vector3(20, 0.5, 20)

[sub_resource type="Environment" id="WorldEnvironment"]
background_mode = 1
background_color = Color(0.12, 0.18, 0.3, 1)
ambient_light_energy = 0.6

[node name="Main" type="Node3D"]

[node name="Player" type="CharacterBody3D" parent="."]
position = Vector3(0, 1, 0)
script = ExtResource("1_controller")
character_manifest_path = "res://characters/imported/character.json"

[node name="CollisionShape3D" type="CollisionShape3D" parent="Player"]
shape = SubResource("PlayerCapsule")

[node name="CharacterRotationRoot" type="Node3D" parent="Player"]

[node name="Character" parent="Player/CharacterRotationRoot" instance=ExtResource("2_character")]

[node name="CameraYaw" type="Node3D" parent="Player"]

[node name="CameraPitch" type="Node3D" parent="Player/CameraYaw"]
rotation = Vector3(-0.349066, 0, 0)

[node name="SpringArm3D" type="SpringArm3D" parent="Player/CameraYaw/CameraPitch"]
spring_length = 5.0

[node name="Camera3D" type="Camera3D" parent="Player/CameraYaw/CameraPitch/SpringArm3D"]
current = true

[node name="Ground" type="StaticBody3D" parent="."]
position = Vector3(0, -0.25, 0)

[node name="MeshInstance3D" type="MeshInstance3D" parent="Ground"]
mesh = SubResource("GroundMesh")

[node name="CollisionShape3D" type="CollisionShape3D" parent="Ground"]
shape = SubResource("GroundShape")

[node name="DirectionalLight3D" type="DirectionalLight3D" parent="."]
rotation_degrees = Vector3(-55, -30, 0)
shadow_enabled = true

[node name="WorldEnvironment" type="WorldEnvironment" parent="."]
environment = SubResource("WorldEnvironment")

Every nested parent is the complete path from the scene root, excluding the root node's own name. Never shorten parent="Player/CameraYaw/CameraPitch" to parent="CameraPitch". A PackedScene instance belongs in the [node ...] header and has no type; never write a typed Node3D followed by instance = ExtResource(...) in the node body. The playable body is a child of the world root. Ground, level geometry, lights, and environment remain siblings of Player, never children of it, so they do not move with the player.

Set character_manifest_path to the imported package's character.json. Bind move_left, move_right, move_forward, move_back, jump, and sprint before running, one summer_input_map_bind per action (e.g. summer_input_map_bind(name="move_forward", events=[{"type": "key", "key": "W"}])); ui_cancel is a built-in action — confirm it is still bound. The controller asserts every one of them at _ready. Do not attach this template during generic import or to an NPC unless the user explicitly asks for player-style control.

The default camera has horizontal mouse orbit only. CameraPitch is set once to a modest downward angle and vertical mouse motion does not change it. The controller keeps a short floor snap plus a small downward grounded velocity so ordinary horizontal movement does not create a false airborne Jump state. It also records the spawn point and respawns after falling below respawn_below_y.

A bind T/A-pose survives every static check: the AnimationPlayer is wired, the clip is named Idle, and the files on disk all read correctly. Only a running frame separates that from a character actually at rest.

Runtime proof. Read this skill's references/default-third-person-verification-probe.gd and pass its full text unchanged as the probe_source of a RunVerification op, which runs it in a hidden, disposable game instance that never touches the editor:

summer_batch(ops=[{"op": "RunVerification", "max_seconds": 30, "probe_source": "<the probe file, unchanged>"}])

The probe samples the live Skeleton3D after the player reaches the floor, requires the real resting clip to be named Idle, verifies that its pose actually changes, rejects an idle whose hands remain level with the upper arms like a bind T/A-pose, saves separate idle and moving frames, verifies W moves toward world -Z, and verifies vertical mouse input leaves camera pitch unchanged. Every boolean in results must be true; open the saved idle and moving frames listed in frames before you describe them. A false is a finding to fix, not a flaky probe.

If the probe cannot run (the op returns ok: false with a failure_reason such as spawn_failed, or the engine build lacks the op), the runtime question belongs to the user. Do not paste the probe into the project. Finish on the saved state: the manifest path written into the Player node, the four semantic roles present, CameraYaw at identity and CameraPitch at its fixed angle in the .tscn text, the InputMap actions bound, and clean diagnostics. Then tell the user which idle clip should be playing at rest, so it is the first thing they look at on Play. That is a complete integration handed off for a playtest, not an incomplete one.

Order of operations

  1. Root node must be CharacterBody2D or CharacterBody3D. move_and_slide() only exists on these. RigidBody is for physics props, not players.
  2. Give it a CollisionShape2D/3D with a real shape (CapsuleShape for 3D humans, RectangleShape/CapsuleShape2D for 2D). No shape = falls through everything.
  3. Read gravity from project settings, never hardcode it, so tuning one value fixes the whole game: ProjectSettings.get_setting("physics/2d/default_gravity").
  4. Bind input actions BEFORE the script references them (move_left, jump, ...). Referencing an unbound action is a silent no-op that reads as "controls dead".
  5. Compile-check, then run and watch it move. Feel is a user call, not a still frame.

2D side-scroller controller

gdscript
extends CharacterBody2D

@export var speed: float = 200.0
@export var jump_velocity: float = -350.0   # negative = up in 2D

var gravity: float = ProjectSettings.get_setting("physics/2d/default_gravity")

func _physics_process(delta: float) -> void:
	if not is_on_floor():
		velocity.y += gravity * delta
	if Input.is_action_just_pressed("jump") and is_on_floor():
		velocity.y = jump_velocity
	var dir := Input.get_axis("move_left", "move_right")
	velocity.x = dir * speed if dir else move_toward(velocity.x, 0.0, speed)
	move_and_slide()

Flip the sprite with $Sprite2D.flip_h = velocity.x < 0 only when velocity.x != 0. For real platformer feel (coyote time, jump buffer, variable height) load references/platformer-feel.md.

Top-down controller (RPG / twin-stick / survivors)

No gravity. 8-directional, normalize so diagonals are not faster.

gdscript
extends CharacterBody2D

@export var speed: float = 220.0

func _physics_process(_delta: float) -> void:
	var dir := Input.get_vector("move_left", "move_right", "move_up", "move_down")
	velocity = dir * speed
	move_and_slide()

Survivors-style games with 200+ agents: enemies are Area2D moved with position += dir * speed * delta, NOT physics bodies. See the make-game skill's survivors blueprint (summer_docs_read id skill/make-game/references/2d-survivors.md). The PLAYER stays a CharacterBody2D.

3D controller (first + third person)

Both share the same body + gravity + Input.get_vector floor logic. The only difference is where the camera lives and how look rotation is applied. Full first-person and third-person controllers, mouse look, sprint, crouch, and spring-arm setup are in references/movement-3d.md - load it for any 3D player.

Minimal 3D walk (no look yet):

gdscript
extends CharacterBody3D

@export var speed: float = 5.0
@export var jump_velocity: float = 4.5
var gravity: float = ProjectSettings.get_setting("physics/3d/default_gravity")

func _physics_process(delta: float) -> void:
	if not is_on_floor():
		velocity.y -= gravity * delta
	if Input.is_action_just_pressed("jump") and is_on_floor():
		velocity.y = jump_velocity
	var input_dir := Input.get_vector("move_left", "move_right", "move_forward", "move_back")
	var direction := (transform.basis * Vector3(input_dir.x, 0.0, input_dir.y)).normalized()
	if direction:
		velocity.x = direction.x * speed
		velocity.z = direction.z * speed
	else:
		velocity.x = move_toward(velocity.x, 0.0, speed)
		velocity.z = move_toward(velocity.z, 0.0, speed)
	move_and_slide()

Summer humanoid package

When the player visual is a Summer character package:

  1. Produce or import the package first — it comes from the shared character pipeline (character-model; assets.generate_3d with rig and animationNames) and carries character.tscn plus its character.json manifest. Instantiate character.tscn; for clip wiring without this controller see character-animation-wiring.
  2. Put the normalized wrapper under a CharacterBody3D; do not modify the package or scale the body.
  3. Let controller velocity and move_and_slide() own all scene translation. Idle, Walk, Run, and Jump are in-place visual clips; never add their root motion to velocity.
  4. Resolve the four clips from character.json semantic roles instead of guessing names. Require every requested role; never substitute one role for another.
  5. Select one state from physics: airborne -> Jump, stopped -> Idle, moving -> Walk, sprinting -> Run. Airborne Jump has highest priority and never falls back to Walk/Run. Change clips only when the state changes, so Jump plays once per takeoff.
  6. Rotate the visual child toward non-zero movement while preserving the normalized -Z forward axis. Turn only visual yaw; do not rotate the collision body merely to turn the mesh. Keep the third-person camera behind the visual on local +Z.

Do not put player controls into the generation pipeline or generic character importer.

Camera-follow basics

  • 2D: add a Camera2D as a CHILD of the player (it is active while enabled, the default), position_smoothing_enabled = true, speed 5-8. Add limit_left/right/top/bottom so it never shows past the level edge. Full rigs in the camera-rigs skill.
  • 3D first-person: camera is a child of a Head node; yaw rotates the body, pitch rotates the Head (never the body) to avoid gimbal lock.
  • 3D third-person: camera on a SpringArm3D so walls push it in. See camera-rigs.

Traps

  • Scaling the CharacterBody node breaks collision + velocity. Scale the MESH/Sprite child instead. In 3D, 1 unit = 1 meter; a human is ~1.0-2.0 tall.
  • is_on_floor() is only valid AFTER move_and_slide() ran once this frame.
  • Diagonal speed-up = you forgot to normalize the input vector.
  • Mouse captured with no release traps the user. Always pair Input.MOUSE_MODE_CAPTURED with a ui_cancel handler that sets it VISIBLE.

From a shell

No Summer MCP tools in this session? Run each one as a command; when summer --version is missing or below 4.0, use npx -y summer-engine@latest in place of summer. Fields are flags in kebab-case (--scene-path res://main.tscn) or one --input '<json>', and summer ops describe <op> prints the exact input. The summer-cli skill has the rest.

In this skillCommand
assets.generate_3dsummer assets generate-3d
summer_docs_readsummer docs read
summer_add_nodesummer engine add-node
summer_batchsummer engine batch
summer_get_diagnosticssummer engine get-diagnostics
summer_get_scene_treesummer engine get-scene-tree
summer_get_script_errorssummer engine get-script-errors
summer_input_map_bindsummer engine input-map-bind
summer_playsummer engine play
summer_read_filesummer engine read-file
summer_replace_textsummer engine replace-text
summer_save_scenesummer engine save-scene
summer_set_propsummer engine set-prop
summer_stopsummer engine stop
summer_write_filesummer engine write-file

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Character Movement AI skill do?

Player controllers for 2D and 3D, plus the standard third-person controller for Summer humanoids.

Why use Character Movement on TypingMind?

Because you install it once and use it with any model. Character Movement is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Character Movement in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/SummerEngine/summer/tree/main/library/skills/character-movement. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Character Movement?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Character Movement?

As many as you like. As long as a model supports skills, you can use Character Movement with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Character Movement AI skill free?

Yes. It is published on GitHub by SummerEngine under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇