agentleFS
szlayszlay/roblox-base/.claude/skills/roblox-studio-mcp/SKILL.md

roblox-studio-mcp

Skill1 starsChanged 4 months agoMITView on GitHub

Use when working with Roblox Studio through built-in MCP for scripts, scenes, generated assets, input, or playtesting.

3 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 2 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    Use the actual exposed schema.

    line 25
  2. Very specific:

    Assets: search and insert existing assets.

    line 30
  3. Distinctive:

    Read before write and read back after every script, asset, or geometry mutation.

    line 44
  4. Very specific:

    Treat generation as asynchronous when a generation ID is returned.

    line 46
  5. Very specific:

    Keep large scripts in multi_edit chunks, not one oversized execution payload.

    line 47
  6. Distinctive:

    If MCP is absent or a capability is unavailable, provide complete offline Luau and state what was not verified.

    line 48

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “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

Roblox Studio MCP

When to Load

Load when working with Roblox Studio through its MCP server: inspecting or editing scripts, building maps and props, generating or inserting assets, debugging, or playtesting. Skip for standalone code generation with no Studio connection.

Quick Reference

Bootstrap before mutation

  1. list_roblox_studios and identify the target.
  2. set_active_studio with the returned ID when selection matters.
  3. get_studio_state and confirm Edit/Client/Server availability.
  4. Inspect the target tree and scripts before changing them.

Use the actual exposed schema. Pass datamodel_type only where the tool requires it: Edit for edit-time changes, Client or Server for runtime operations. Do not guess from a previous session.

Capabilities

  • Inspect: search_game_tree, inspect_instance, script_search, script_read, script_grep
  • Edit/execute: multi_edit, execute_luau
  • Assets: search and insert existing assets; generate_mesh, generate_material, generate_procedural_model, wait_job_finished, store_image, upload_image
  • Play/evidence: start_stop_play, get_console_output, screen_capture, character_navigation, user_keyboard_input, user_mouse_input, subagent

The docs use names such as search_asset and insert_asset; other bridges may expose aliases such as search_creator_store and insert_from_creator_store. Route by capability, then inspect the live schema.

Execution contract

discover → select Studio/context → inspect → mutate in bounded batches
→ read back → playtest when needed → inspect console/visual evidence
→ clean up or report the exact fallback

Reliability rules

  • execute_luau is stateless. Re-acquire references every call.
  • Read before write and read back after every script, asset, or geometry mutation.
  • Generate or insert assets only after choosing between reuse, procedural generation, mesh/material generation, and native fallback.
  • Treat generation as asynchronous when a generation ID is returned; call wait_job_finished with that generationId before dependent edits.
  • Keep large scripts in multi_edit chunks, not one oversized execution payload.
  • If MCP is absent or a capability is unavailable, provide complete offline Luau and state what was not verified.

Screenshot cost

screen_capture has no resolution parameter and its image cannot be downscaled after the fact - it arrives in context as it is. The Studio window size is the only knob:

WindowViewportCost per capture
maximized (2048 wide)~1568x860~1800 tokens
-Medium (1600x1000)~980x635~830 tokens
-Small (1100x720)~370x400~200 tokens

A PreToolUse hook (scripts/hooks/pre-capture-resize.ps1) already sizes the window before every capture, and no-ops when it is already right - do not resize by hand around a capture. For one shot that needs detail a thumbnail cannot carry, put full (or small) in the capture_id: ScreenCapture_hud_full. It applies to that call only and the next unmarked capture reverts on its own, so there is no mode to remember to switch off.

Cheaper still is not capturing: inspect_instance and execute_luau return positions, sizes, colours and visibility as text for a fraction of the cost. Capture to judge how it looks, not to confirm that it exists.

Full tool mappings, live-schema differences, asset workflows, evidence recipes, and recovery rules: references/full.md

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.