Skip to content

andrian-syh/roblox-optimum

v1.19.1MIT

Professional Roblox and Luau standards for AI agents: server-authoritative, leak-free, framework-agnostic, and laid out the same way in every file.

best-practices

Roblox/Luau coding standards for writing any game code. Use whenever the user asks to write, add, build, implement, fix, or refactor Luau (Script, LocalScript, ModuleScript) or a game system in a Roblox project - a shop, DataStore save, RemoteEvent, NPC spawner, UI - even when Roblox or Luau is never named, and when the user asks to keep best practices in mind as standing guidance. Enforces the VARIABLES/FUNCTIONS/INITIALIZATION layout, naming, performance, memory, networking, and security rules under any framework or genre, adapting to the project's style and community libraries (ProfileStore, Trove, Knit, Fusion, ...). Hand-offs: reviewing or auditing existing code goes to the code-review skill, a reported symptom with no file named to the diagnose skill, sync, Studio MCP, or playtests to the studio-ops skill. Not for non-Roblox Lua; launch and marketing with no Luau go to the publishing skill.

code-review

Reviews and audits existing Roblox/Luau code. Use whenever the user asks to review, check, or audit a file, diff, pull request, place, or system, asks whether their code is safe, good, or exploitable, or wants its architectural health or maturity scored - even when Roblox or Luau is never named but the code is a game script. Finds leaks, exploits, and data-loss risks, reports one severity per finding (Blocker / Correctness / Advisory) behind a confidence gate, and scores one system 1-5 across security, lifecycle, CPU, network, data safety, and structure. Hand-offs: writing or refactoring code goes to the best-practices skill, chasing a reported symptom with no file named to the diagnose skill, Studio tooling, sync, or playtest questions to the studio-ops skill, store listings and launch plans to the publishing skill.

diagnose

Diagnoses a reported Roblox symptom down to its cause before anything is changed. Use whenever the user says something in their game is broken, wrong, missing, duplicated, intermittent, or works in Studio but not live - leaderstats resetting, players falling through the map, a shop granting free items, a server dying after a while, an event firing twice - and no file has been named yet, even when Roblox is never mentioned. Classifies the symptom, asks which side owns the state and whether a client could have caused it, narrows the surface, reproduces, and confirms the cause before handing the fix over. Hand-offs: judging code that was handed over goes to the code-review skill, writing the fix to the best-practices skill, sync toolchains and driving Studio tools to the studio-ops skill, falling impressions or player counts with nothing broken to the publishing skill.

publishing

Plans and checks how a Roblox experience reaches players, outside the code: store listing, icon, thumbnails and video, Recommended for You and the other discovery surfaces, the maturity questionnaire and Roblox Kids and Select age tiers, launch readiness, Ads Manager, experience events and notifications, reading retention and revenue analytics, update cadence, and the marketing Roblox forbids. Use whenever the user asks how to launch, promote, market, or grow a game, why impressions or players fell, what a title or thumbnail should be, or whether a growth idea breaks platform policy, even when no code is involved. Hand-offs: writing or changing Luau, including analytics events and odds disclosure, goes to the best-practices skill, judging existing code to the code-review skill, broken game behaviour to the diagnose skill, Studio tooling, sync, and playtests to the studio-ops skill.

studio-ops

Operates a Roblox project's tooling safely: Studio MCP connections, filesystem sync with Rojo, Argon, Script Sync, or Azul, and proving a change works in the running engine. Use before the first Studio MCP tool call of a session (list_roblox_studios, execute_luau, multi_edit, start_stop_play, and the rest), and whenever the user asks why their edits are being overwritten, how to set up or fix a sync toolchain, what a Studio MCP tool will do before it runs, how to playtest or verify a fix, or how to test replication and multi-client behaviour. Governs how the agent drives its tools, not how Luau is written. Hand-offs: authoring Luau goes to the best-practices skill, judging existing code to the code-review skill, working out why a reported symptom happens to the diagnose skill, which calls on this one to run the probe, launch and marketing to the publishing skill.