agentleFS
yurineko73yurineko73/Godot-MCP-Native/skills/gdmcp/SKILL.md

gdmcp

Skill836 starsChanged 6 months agoMITView on GitHub

Use the installed gdmcp CLI to inspect, edit, run, and debug a Godot project without loading the complete Godot MCP tool catalog.

The file

Show its rules and text

We split this skill into the rules it gives an agent

14 rules. About 0 are common, 4 are distinctive, and 10 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 10

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use gdmcp from the shell for operations that require the running Godot editor.

    line 8
  2. Very specific:

    Run the CLI from the Godot project root when possible.

    line 13
  3. Distinctive:

    When the current directory is not the project root, provide the project and, if necessary, Godot's custom user-data root explicitly:

    line 19
  4. Distinctive:

    Prefer narrow domain commands for common operations:

    line 39
  5. Distinctive:

    When a domain command is unavailable, use progressive discovery:

    line 75
  6. Very specific:

    Use --json when the output will be analyzed programmatically.

    line 85
  7. Very specific:

    Prefer domain commands over raw tool-call.

    line 86
  8. Distinctive:

    Do not request the complete catalog with full schemas.

    line 87
  9. Very specific:

    Use scripts list --limit <n> [--cursor <cursor>] for progressive script discovery.

    line 89
  10. Very specific:

    Use scripts read --lines <start>:<end> to read specific line ranges.

    line 90
  11. Very specific:

    Use nodes get --fields <field1,field2> and resources get --fields <field1,field2> to retrieve only specific properties.

    line 91
  12. Very specific:

    Use {scenes,nodes,scripts,resources} resolve <name> to convert human-readable names to stable paths.

    line 92
  13. Very specific:

    Use debug logs --cursor <offset> to continue log pages.

    line 94
  14. Very specific:

    Use batch preview before batch apply.

    line 96

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

gdmcp

Use gdmcp from the shell for operations that require the running Godot editor. By default, invoke the project-local CLI from the project root at .\.gdmcp\bin\gdmcp.exe in PowerShell; the gdmcp command in the examples below is shorthand for this path. Use a PATH-installed gdmcp only as an explicit fallback when the local executable is unavailable.

Project directory and configuration discovery

Run the CLI from the Godot project root when possible. The CLI uses the process working directory to discover project.godot, then resolves the project's Godot user://mcp_settings.cfg so a persisted HTTP http_port is used automatically. The CLI executable directory (for example .gdmcp/bin) is not used as the project directory.

When the current directory is not the project root, provide the project and, if necessary, Godot's custom user-data root explicitly:

gdmcp --project-path <project-root> --godot-user-data-dir <godot-user-data-root> --json doctor

Equivalent environment variables are GODOT_MCP_PROJECT_PATH and GODOT_USER_DATA_DIR. URL precedence remains: explicit --url, GODOT_MCP_URL, the CLI config file, the discovered Godot MCP settings, then the built-in http://127.0.0.1:9080 fallback. Stdio MCP settings do not produce an HTTP endpoint and therefore use the normal fallback chain.

Start with:

gdmcp --json doctor
gdmcp --json editor state

Prefer narrow domain commands for common operations:

gdmcp --json scenes current
gdmcp --json scenes list --limit 20
gdmcp --json scenes tree --depth 4
gdmcp --json nodes list --limit 20
gdmcp --json nodes get /root/Main/Player --fields position,visible
gdmcp --json nodes properties set /root/Main/Player --property speed --value 300
gdmcp --json scripts list --limit 20
gdmcp --json scripts read res://player.gd --lines 1:200
gdmcp --json scripts create res://new_script.gd --script-type GDScript
gdmcp --json resources list --limit 20
gdmcp --json resources get res://player.tres --fields resource_path
gdmcp --json project settings --filter display/
gdmcp --json debug logs --limit 50
gdmcp --json runtime tree --depth 4
gdmcp --json runtime nodes get /root/Main/Player

Resolve names to stable paths (no tool-call needed):

gdmcp --json nodes resolve Player
gdmcp --json scenes resolve Main
gdmcp --json scripts resolve player
gdmcp --json resources resolve icon

Refactor nodes:

gdmcp --json nodes move /root/Main/Player --new-parent /root/World
gdmcp --json nodes rename /root/Main/Enemy1 --new-name Boss

When a domain command is unavailable, use progressive discovery:

gdmcp --json tools search "<intent>" --limit 5
gdmcp --json tools schema <tool-name>
gdmcp --json tool-call <tool-name> --args-file <request.json>

Rules:

  • Use --json when the output will be analyzed programmatically.
  • Prefer domain commands over raw tool-call.
  • Do not request the complete catalog with full schemas.
  • Bound output with --limit, --depth, --fields, --lines, --max-bytes, or --out.
  • Use scripts list --limit <n> [--cursor <cursor>] for progressive script discovery.
  • Use scripts read --lines <start>:<end> to read specific line ranges.
  • Use nodes get --fields <field1,field2> and resources get --fields <field1,field2> to retrieve only specific properties.
  • Use {scenes,nodes,scripts,resources} resolve <name> to convert human-readable names to stable paths.
  • scenes list, nodes list, scripts list, resources list, and debug logs default to 50 items.
  • Use debug logs --cursor <offset> to continue log pages.
  • Always pass project settings --filter <prefix>.
  • Use batch preview before batch apply.
  • Batch files use registered tool names and object arguments. Validation completes before any request is sent; execution is sequential and non-atomic.
  • Destructive domain commands require --apply.
  • Raw destructive calls require --apply after reviewing the schema.
  • Runtime and open-world tools require --allow-open-world.
  • Configure tokens via GODOT_MCP_TOKEN environment variable; never print them.

See references/command-workflows.md for copyable task flows.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.