Agent Native

unity-editor

Control the Unity Editor to enter and exit play mode, step through frames, inspect runtime state, manage selection, and execute menu commands.

by Besty0728CommunityNot reviewed

Community author

Agents
Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode, claude.ai and API
License
MIT (permissive)
Pin
15a1925pinned Oct 9, 2026
Context cost
about 3,300 tokenswhen the skill loads
GitHub stars
1,819on besty0728/unity-skills
Scripts
None found

What it does

“Control and observe the Unity Editor state”
The author's description, from SKILL.md

Generated summary from the skill's own text

Control the Unity Editor to enter and exit play mode, step through frames, inspect runtime state, manage selection, and execute menu commands. The skill reads scene changes without parsing YAML and provides observation of editor and play mode state.

When to use it

  • Entering or exiting play mode to test runtime behavior
  • Inspecting GameObject transform and component state during gameplay
  • Stepping through frames to debug frame-by-frame behavior
  • Executing menu items like File/Save or Assets/Refresh

When not to

  • Modifying scenes during play mode, as changes are lost on exit
  • Compiling code or checking compilation status, use debug module instead
  • Parsing raw scene YAML directly, use scene module instead

Install

These commands install from the publisher's repository and need no account.

skills CLI

npx skills add besty0728/unity-skills --skill unity-editor

Installs from the repository's default branch, which may have changed since the commit we checked.

GitHub CLI 2.90 or later

gh skill install besty0728/unity-skills unity-editor@15a1925bcbeebdd4282589c6464f6e2f02041401

Installs the commit we checked (15a1925).

Copy by hand

Download the folder at 15a1925 and put it in .claude/skills/unity-editor for Claude Code and in .agents/skills/unity-editor for Codex, Cursor, Gemini CLI, GitHub Copilot and OpenCode.

Skill folder on GitHub at 15a1925

Copy for your agent

A prompt for any coding agent. It installs this skill from GitHub at the pinned commit, copies the files as they are, and runs nothing unless you ask.

Prompt

Install the agent skill "unity-editor" by Besty0728.

Source, pinned to commit 15a1925bcbeebdd4282589c6464f6e2f02041401:
https://github.com/besty0728/unity-skills/tree/15a1925bcbeebdd4282589c6464f6e2f02041401/SkillsForUnity/unity-skills~/skills/editor

1. Download that folder at exactly this commit. Do not use the default branch.
2. Copy every file unchanged, keeping the folder structure, into .claude/skills/unity-editor/ for Claude Code and .agents/skills/unity-editor/ for Codex, Cursor, Gemini CLI, GitHub Copilot and OpenCode. Create the folders if they do not exist.
3. Do not edit the files, and do not run any script in the folder unless I ask you to.
4. When you are done, list the files you copied and where you put them.

Listing and checks: https://www.agentnative.dev/skills/besty0728/unity-editor

SKILL.md at the pin

---
name: unity-editor
description: Control and observe the Unity Editor state
---

> **Before calling any skill in this module:** if you are about to call a skill with parameters guessed from its name or description, STOP — read this file (or fetch its schema via `GET /skills/recommend?includeSchema=true`) first. If you already have the parameter definitions from recommend/schema, you may proceed straight to dryRun.

## Triggers
- Reading persistent scene/file changes
- Driving play mode
- Inspecting runtime state
- Executing menu items
- 读取持久化场景/文件变更、操控 Play Mode、检查运行时状态、执行菜单项

# Unity Editor Skills

Observe and control the Unity Editor without parsing scene YAML.

## Operating Mode

- **Approval**:本模块 Mixed —— `editor_get_changes` / `editor_get_selection` / `editor_get_context` / `editor_get_state` / `editor_get_tags` / `editor_get_layers` / `editor_playmode_inspect` 标 `SkillMode.SemiAuto`,可直接执行;其余 `editor_select` / `editor_undo` / `editor_redo` / `editor_playmode_step` 默认 FullAuto,Approval 模式下需 grant。
- **Auto / Bypass**:FullAuto 直接执行。
- **含 NeverInSemi 高危 skill**:`editor_play` / `editor_play_capture` / `editor_stop` / `editor_pause`(标 `MayEnterPlayMode = true`);`editor_execute_menu`(标 `MayTriggerReload = true` —— `Assets/Refresh`、`Assets/Reimport All` 之类的菜单项会触发域重载)。这些在 Approval **和** Auto 下都返 `MODE_FORBIDDEN`,仅 Bypass 或 Allowlist 命中可调;**grant 流程对它们不适用**,不要尝试。

**DO NOT** (common hallucinations):
- `editor_run` does not exist → use `editor_play` to enter play mode
- `editor_compile` / `editor_recompile` do not exist → use `debug_force_recompile`
- `editor_save` does not exist → use `editor_execute_menu` with menuPath `"File/Save"`
- `editor_execute_menu` requires exact menu path — typos cause silent failure

**Routing**:
- For compilation check → use `debug` module's `debug_check_compilation`
- For console errors → use `debug` module's `debug_get_errors`
- For scene save → `scene_save` (scene module) or `editor_execute_menu` menuPath="File/Save"
- When file watching reports changes, or the AI resumes after the user edited Unity → call `editor_get_changes` before reading `.unity` YAML
- If `editor_get_changes.dropped=true` → its 500-entry retention window was exceeded; rebuild context with `scene_context` / `scene_diff`

## Skills Overview

| Skill | Description |
|-------|-------------|
| `editor_play` | Enter play mode |
| `editor_play_capture` | Observe runtime errors, optionally screenshot, then exit |
| `editor_stop` | Exit play mode |
| `editor_pause` | Toggle pause |
| `editor_playmode_step` | Advance Play Mode by N frames (async job) |
| `editor_playmode_inspect` | Inspect a GameObject's live runtime state (transform, component fields) |
| `editor_select` | Select GameObject |
| `editor_get_selection` | Get selected objects |
| `editor_get_context` | Get full editor context (selection, assets, scene) |
| `editor_get_changes` | Read persistent scene/file changes by cursor |
| `editor_undo` | Undo last action |
| `editor_redo` | Redo last action |
| `editor_get_state` | Get editor state |
| `editor_execute_menu` | Execute menu item |
| `editor_get_tags` | Get all tags |
| `editor_get_layers` | Get all layers |
| `console_set_pause_on_error` | Pause play mode on error (console module) |

---

## Skills

### editor_play
Enter play mode. Warning: any unsaved scene changes made during Play mode will be lost when exiting.

**Returns**: `{success, mode, jobId}` — `mode="playing"`, `jobId` returned from `AsyncJobService` so callers can poll `entering_play_mode` completion.

### editor_play_capture
Enter Play Mode, observe errors for `durationSeconds` (default 10, range 1–300), optionally capture the Game View, then exit. Returns a Job whose result includes `healthy`, error aggregates, `stoppedEarly`, and `screenshotPath`.

### editor_stop
Exit play mode.

**Returns**: `{success, mode}` — `mode="stopped"`.

### editor_pause
Toggle pause state.

**Returns**: `{success, paused}` — `paused` is the new boolean state.

### editor_playmode_step
Advance Play Mode forward by `frames` (1–100, default 1) using `EditorApplication.Step`; Unity automatically enters paused state as part of stepping. Requires Play Mode to already be active — call `editor_play` or `editor_play_capture` first, otherwise this returns a structured error (`error` + `hint` + `suggestedSkills: ["editor_play", "editor_play_capture"]`). Only one step job may be in flight at a time; a second call while one is still running returns an error naming the active `jobId`.

| Parameter | Type | Required | Default | Description |
|-----------|------|----------|---------|-------------|
| `frames` | int | No | 1 | Frames to advance, clamped to 1–100 |

**Runs as an async job, not synchronously**: `EditorApplication.Step()` only lands on a later Editor tick, so the skill issues one `Step()` call at a time, confirms it landed by watching `Time.frameCount` advance, and only then issues the next — back-to-back `Step()` calls without that confirmation are not reliable. The call itself returns immediately: `{success, status: "accepted", jobId, framesRequested}`.

Poll `job_status` with the returned `jobId` until `status="completed"`; its `details` then contains `{framesRequested, framesCompleted, frameCount, isPaused}`. **Do not use `job_wait`** — like `playmode`/`play_capture`, this job's progress depends on `EditorApplication.update` ticks, and `job_wait`'s blocking loop runs on the same main thread those ticks come from, so it cannot observe progress; it will just spend its full timeout doing nothing. If Play Mode exits or Unity fails to advance a frame within 10s, the job fails with `failed_exited_play_mode` / `failed_step_timeout`.

### editor_playmode_inspect
Inspect a GameObject's live runtime state: transform (position/rotation/scale), `activeSelf`/`activeInHierarchy`, and — when `componentType` is given — that component's public fields and properties (reuses the same reflection helper as `component_get_properties`, not reimplemented). Works during Play Mode, including while paused, and also in Edit Mode, where it returns editor-time values — check `isPlaying`/`isPaused` in the response to know which state the values reflect.

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| `name` | string | No* | Object name |
| `instanceId` | int | No* | Instance ID (preferred) |
| `path` | string | No* | Object path |
| `componentType` | string | No | Component type name; when set, also returns its public fields/properties |

*One identifier required

**Returns**: `{success, gameObject, entityId, instanceId, path, activeSelf, activeInHierarchy, transform: {position, localPosition, rotation, localScale}, isPlaying, isPaused, component}`. `component` is `null` when `componentType` is omitted, an error object when the type isn't found or isn't attached, or `{gameObject, component, fullTypeName, properties, fields}` (same shape as `component_get_properties`) otherwise.

### editor_select
Select a GameObject.

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| `name` | string | No* | Object name |
| `instanceId` | int | No* | Instance ID (preferred) |
| `path` | string | No* | Object path |

*One identifier required

### editor_get_selection
Get currently selected objects.

**Returns**: `{count, objects: [{name, instanceId}]}`

### editor_get_context
Get full editor context including selection, assets, and scene info.

| Parameter | Type | Required | Default | Description |
|-----------|------|----------|---------|-------------|
| `includeComponents` | bool | No | false | Include component list |
| `includeChildren` | bool | No | false | Include children info |

**Returns**:
- `selectedGameObjects`: Objects in Hierarchy (instanceId, path, tag, layer)
- `selectedAssets`: Assets in Project window (GUID, path, type, isFolder)
- `activeScene`: Current scene info (name, path, isDirty)
- `focusedWindow`: Name of focused editor window
- `isPlaying`, `isCompiling`: Editor state

### editor_get_changes
Read the persistent change journal at `Library/UnitySkills/editor_changes.jsonl`. It captures scene object/component/property summaries plus imported, deleted, and moved asset paths across Domain Reloads. It never parses or returns raw `.unity` YAML.

| Parameter | Type | Required | Default | Description |
|-----------|------|----------|---------|-------------|
| `since` | long | No | `0` | Previous response cursor; `0` reads retained history |
| `types` | string | No | `all` | Comma-separated `scene`, `file`, `undo`, or `lifecycle` |
| `source` | string | No | `all` | `editor`/`manual`, `rest`, or `all` |
| `limit` | int | No | `100` | Newest entries to return, 1-500 |

**Returns:** `{hasChanges, cursor, oldestSeq, dropped, truncated, changes}`. Save `cursor` for the next call. `dropped=true` means older changes fell outside the 500-entry journal, so rebuild full scene context.

### editor_undo
Undo the last action.

### editor_redo
Redo the last undone action.

### editor_get_state
Get current editor state.

**Returns**: `{isPlaying, isPaused, isCompiling, timeSinceStartup, unityVersion, platform}`

### editor_execute_menu
Execute a menu command.

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| `menuPath` | string | Yes | Menu item path |

**Common Menu Paths**:
| Menu Path | Action |
|-----------|--------|
| `File/Save` | Save current scene |
| `File/Build Settings...` | Open build settings |
| `Edit/Play` | Toggle play mode |
| `GameObject/Create Empty` | Create empty object |
| `Window/General/Console` | Open console |
| `Assets/Refresh` | Refresh assets |

### editor_get_tags
Get all available tags.

**Returns**: `{tags: [string]}`

### editor_get_layers
Get all available layers.

**Returns**: `{layers: [{index, name}]}`

### Pause On Error
Pause-on-error is provided by the console module, not the editor module.

Use `console_set_pause_on_error` from [console/SKILL.md](../console/SKILL.md).

---

## Example Usage

```python
import unity_skills

# Check editor state before operations
state = unity_skills.call_skill("editor_get_state")
if state['isCompiling']:
    print("Wait for compilation to finish")

# On resume or after file-watch notification, inspect semantic changes first.
changes = unity_skills.call_skill("editor_get_changes", since=last_cursor)
last_cursor = changes["cursor"]

# Get full context (useful for understanding current state)
context = unity_skills.call_skill("editor_get_context", includeComponents=True)
for obj in context['selectedGameObjects']:
    print(f"Selected: {obj['name']} (ID: {obj['instanceId']})")

# Select and operate on object
unity_skills.call_skill("editor_select", name="Player")
selection = unity_skills.call_skill("editor_get_selection")

# Safe experimentation with undo
unity_skills.call_skill("gameobject_delete", name="TestObject")
unity_skills.call_skill("editor_undo")  # Restore if needed

# Execute menu command
unity_skills.call_skill("editor_execute_menu", menuPath="File/Save")

# Step through Play Mode frame-by-frame and assert state changed
import time

# Enter Play Mode with a generous observation window so there's time left to step
capture = unity_skills.call_skill("editor_play_capture", durationSeconds=30)
while True:
    capture_status = unity_skills.call_skill("job_status", jobId=capture["jobId"])
    if capture_status["currentStage"] not in ("entering_play_mode", "domain_reload_recovery"):
        break  # Play Mode has been entered; capture_status["currentStage"] == "observing"
    time.sleep(0.2)

before = unity_skills.call_skill("editor_playmode_inspect", name="Player")["transform"]["position"]

step = unity_skills.call_skill("editor_playmode_step", frames=3)
while True:
    step_status = unity_skills.call_skill("job_status", jobId=step["jobId"])  # not job_wait, see above
    if step_status["status"] in ("completed", "failed"):
        break
    time.sleep(0.1)

after = unity_skills.call_skill("editor_playmode_inspect", name="Player")["transform"]["position"]
assert after != before, "Player did not move after 3 frames"
print(step_status["details"]["frameCount"], step_status["details"]["isPaused"])
```

## Best Practices

1. Check editor state before play mode operations
2. Don't modify scene during play mode (changes lost)
3. Use undo for safe experimentation
4. On resume, call `editor_get_changes` before reading scene files
5. Use `editor_get_context` to get instanceId for batch operations
6. Menu commands must match exact paths

## Exact Signatures

Exact names, parameters, defaults, and returns are defined by `GET /skills/schema` or `unity_skills.get_skill_schema()`, not by this file.

## Common Errors

Full transport-level codes (COMPILING/RATE_LIMIT etc.) → ../../references/protocol-error-codes.md

| Error | Trigger | Fix |
|---|---|---|
| `TARGET_NOT_FOUND` | The requested GameObject or menu item could not be found (e.g., `Menu item not found or failed`). | Verify the object with `gameobject_find` / `scene_get_hierarchy`, or check the exact menu path spelling before retrying. |
| `SKILL_ERROR` | A play-mode state conflict occurred, such as `Already in play mode`, `Not in play mode`, or an active frame-step job already exists. | Match the editor state to the skill requirement: enter/exit play mode first, or wait for the existing step job to finish. |

From besty0728/unity-skills at 15a1925, shown as plain text under its license: MIT (permissive). The license text is in LICENSE.

View SKILL.md on GitHub at 15a1925

Files

1 file, 14 KB in total. Each file has a sha256 in the pin's manifest; the first 12 characters are shown.

Skill folder on GitHub at 15a1925
All Coding workflow skills
  • planning

    This skill orchestrates multi-workflow projects with dependencies and shared data tables by creating a task graph in the orchestrator via the create-tasks tool.

    by n8nOfficial publisher

    Claude Code, Codex, Cursor +3 · about 2.4K tokens · pinned 6 days ago

  • progressive-building

    This skill implements incremental workflow development for agents building or extending workflows.

    by n8nOfficial publisher

    Claude Code, Codex, Cursor +3 · about 1.4K tokens · pinned 6 days ago

  • workflow-builder

    Builds and edits n8n workflows using the TypeScript SDK.

    by n8nOfficial publisher

    Claude Code, Codex, Cursor +3 · about 14.2K tokens · pinned 6 days ago

  • remotion-upgrade

    This skill upgrades Remotion and its related packages to the latest stable version.

    by RemotionOfficial publisher

    Claude Code, Codex, Cursor +3 · about 460 tokens · pinned 6 days ago

  • remotion-upgrade

    This skill upgrades Remotion and related packages across a project, automatically detecting the package manager and workspaces.

    by RemotionOfficial publisher

    Claude Code, Codex, Cursor +3 · about 460 tokens · pinned 6 days ago

  • final-release-review

    Audits Python SDK release candidates or plans against the previous release, comparing diffs for regressions, API changes, and version compatibility.

    by OpenAIOfficial publisher

    Contains scripts

    Claude Code, Codex, Cursor +3 · about 5.4K tokens · pinned 6 days ago

The skills listed here are third-party software, collected from public GitHub repositories as a catalog of recent skills and provided as is. Agent Native does not write or control them and is not responsible for any harm, loss or damage caused by installing or using them. Labels and checks describe what we saw at one commit and are not a guarantee. Read a skill's files before you run it. Terms of use