m4bwav/roblox-agent
Evergreen skills for building Roblox games with AI coding agents: connecting to Studio's built-in MCP server the safe way, keeping code on disk with Script Sync or Rojo, checking Luau without Studio (build, lint, format, types, Lune tests), Open Cloud Luau Execution and place uploads, and agent rules for secure, current Roblox code. Each skill re-researches its subject on a schedule and carries the evals that prove it triggers and acts.
Verify Roblox Luau code without opening Studio, the way an agent should before saying a change is done: one script (rbx_check.py) runs rojo build, rojo sourcemap, selene, stylua --check, luau-lsp analyze with the Roblox type definitions, and an optional Lune test, and prints one JSON report; plus the verification ladder (static checks, Lune unit tests, place-file assertions with @lune/roblox, engine tests on Roblox servers, a Studio playtest) and a GitHub Actions workflow for a Roblox repo. Use whenever the user asks to 'check my Roblox code', 'lint the Luau', 'type check with luau-lsp', 'does this build', 'set up CI for my Roblox game', 'run the tests', 'why does luau-lsp say Unknown global game', or after the agent edits Luau in a Rojo project, or says 'refresh roblox-agent-checks'. Engine tests through Open Cloud are roblox-agent-cloud; playtests in Studio are roblox-agent-studio.
Use Roblox Open Cloud from an agent without Studio: run Luau headlessly on a real Roblox server (Luau Execution: create a task, poll it, read results and logs), run engine-level Jest Roblox tests that way in CI, and save or publish a place version from a built .rbxl, through one stdlib script (rbx_cloud.py) that reads the API key from the environment and prints one JSON object. Covers key scopes, limits, a dedicated test place, Saved versus Published, and the 2026-10-19 cut to cookie-based web API limits. Use whenever the user asks to 'run Luau on Roblox servers', 'run my tests in CI against the real engine', 'Open Cloud Luau execution', 'publish my place from the command line', 'upload a place version', 'set up an Open Cloud API key', or a script hits rate limits on Roblox web APIs, or says 'refresh roblox-agent-cloud'. Local checks are roblox-agent-checks; Studio work is roblox-agent-studio.
Put a Roblox game's code on disk so an AI coding agent can edit it as ordinary files under Git: choose between Studio's Script Sync and Rojo, set up the Rokit toolchain (rojo, StyLua, selene, luau-lsp, lune, wally or pesde) in a rokit.toml, scaffold or convert a project (rojo init, rojo syncback), keep one writer per script so Studio, sync and MCP edits never fight, and set up luau-lsp for editor types. Use whenever the user asks to 'set up Rojo', 'use Claude Code on my Roblox game from VS Code', 'sync my scripts to files', 'Script Sync or Rojo', 'start a new Roblox project', 'add rokit', 'install wally packages', 'move my place to Git', or reports sync conflicts or lost edits between Studio and disk, or says 'refresh roblox-agent-project'. Running checks and fixing type errors is roblox-agent-checks; connecting to Studio is roblox-agent-studio.
Connect an AI coding agent (Claude Code, Codex, Cursor, Copilot) to a running Roblox Studio through its built-in MCP server and use it safely: Roblox's official plugin first, one manual entry otherwise, never two; prove the connection with list_roblox_studios; run the playtest-and-verify loop (start_stop_play, execute_luau, get_console_output); reach Roblox's Studio skills through the skill tool; work around known bugs (disconnects, the execute_luau sandbox, stale reads before multi_edit). Use whenever the user asks to 'connect Claude to Roblox Studio', 'set up the Roblox MCP', 'let the agent see my place', 'playtest it for me', 'why can't Claude insert scripts', reports the Studio MCP disconnecting or not listing Studio, asks which Roblox MCP server to use, or says 'refresh roblox-agent-studio'. Code on disk is roblox-agent-project; checks without Studio are roblox-agent-checks.
How an AI coding agent writes Roblox code that survives review: look up APIs in Roblox's agent-ready docs (llms.txt, .md pages, the deprecated list) instead of recalling them; avoid the deprecated calls models still emit (wait, BodyVelocity, Humanoid:LoadAnimation); never trust the client (RemoteEvent validation, rate limits); keep DataStores safe (UpdateAsync, pcall, session locks); check Creator Store inserts for backdoors; follow the rules for player-facing AI; write an AGENTS.md for a Roblox repo. Use whenever the agent writes or reviews Luau for a Roblox game, or the user asks 'review my Roblox script', 'is this exploitable', 'make this RemoteEvent secure', 'save player data', 'what's deprecated', 'write an AGENTS.md for my Roblox game', 'add an AI NPC', or says 'refresh roblox-agent-workflow'. Project setup is roblox-agent-project; running checks is roblox-agent-checks.