Is TemporaryLooker22/Stud-UI-Clean-Roblox safe?
Yes. TemporaryLooker22/Stud-UI-Clean-Roblox is safe to install. Oathe's behavioral security audit gave the Stud-UI-Clean-Roblox skill by TemporaryLooker22 a trust score of 92/100 with 5 findings, none critical or high.
https://github.com/TemporaryLooker22/Stud-UI-Clean-Roblox
Is TemporaryLooker22/Stud-UI-Clean-Roblox safe to install?
Stud-UI-Clean-Roblox is a legitimate, well-documented Roblox simulator UI design system skill. The SKILL.md contains no prompt injection attempts, no instructions to access sensitive files, and no exfiltration mechanisms. The only substantive concerns are an aggressively broad skill-trigger description that could activate the skill on generic user requests without explicit invocation, and the inherent code-execution surface created by directing the agent to use Roblox Studio MCP execute_luau with the included Luau source files. All canary honeypots remained intact, the clone connected only to GitHub, and no unexpected filesystem or network activity was observed.
What security issues were found in TemporaryLooker22/Stud-UI-Clean-Roblox?
Category Scores
Findings (5)
LOW Aggressive broad trigger condition in skill description -12 ▶
The SKILL.md description instructs the hosting agent to activate this skill 'even if they only say make a shop menu, add a button to the HUD, or UI simulator style without naming the style.' This greedy matching pattern means the skill will inject its ~130KB context into agent conversations for vague, generic UI requests that the user may not have intended to route through this skill. While not malicious, it hijacks agent judgment about skill applicability.
Evidence: description field: '...even if they only say make a shop menu, add a button to the HUD or UI simulator style without naming the style'
LOW Luau source code included with agent-directed execute_luau pathway -10 ▶
The skill includes three Luau ModuleScript source files and six example builders, and explicitly instructs the agent to run them via Roblox Studio MCP execute_luau. While Luau is sandboxed to Roblox Studio and cannot access the host filesystem, this establishes a code-execution surface inside Roblox Studio. A malicious actor who forked this skill could replace the Luau modules with scripts that call Roblox HTTPService to exfiltrate game data.
Evidence: SKILL.md Workflow step 3: 'run once from the command bar / MCP execute_luau'; src/StudUI.luau, src/StudMotion.luau, src/CheckUIIcons.luau present
INFO Roblox custom asset IDs depend on creator account integrity -7 ▶
Six textures (studs tile rbxassetid://140544932962041, sunburst 130563114903838, shop header 108268925256584, basket 120530633952826, robux coin 83084350609799, rarity panels) are uploaded to Roblox by the skill author. If the creator account were compromised, these assets could be swapped for inappropriate content. This is a Roblox-platform supply chain consideration, not a host-system risk.
Evidence: references/icons-and-assets.md §2 Custom textures table; UI.Assets and UI.Rarities references in SKILL.md
INFO Network activity limited to GitHub clone endpoint -4 ▶
The only external TCP connection made during install was to 140.82.121.4:443 (GitHub.com) for the HTTPS clone. DNS resolved github.com correctly. No connections to unexpected endpoints were observed.
Evidence: Network monitoring: saddr=020001BB8C5279040000000000000000 → 140.82.121.4:443; DNS to 10.0.2.3:53 for github.com
INFO French instance naming convention required by runtime scripts -8 ▶
SKILL.md instructs the agent to preserve French instance names (Fenetre, Entete, Corps, Biseau, etc.) because runtime scripts find UI parts by these names. This is a legitimate architectural requirement from the reference game, not an attempt to obfuscate or encode data.
Evidence: SKILL.md Rules of thumb: 'Keep the French instance names (Fenetre, Entete, Corps...) runtime scripts find parts by these names'
Should I install TemporaryLooker22/Stud-UI-Clean-Roblox?
Oathe's verdict for TemporaryLooker22/Stud-UI-Clean-Roblox is SAFE with a trust score of 92/100. Recommendation: Install.