hishambaoom

Roblox Agent Skill No Readme — Testing skill for Claude Code

Testing community

How to install Roblox Agent Skill No Readme

This entry records only its repository, not the path inside it, so there is no exact command to give. Open hishambaoom/roblox-agent-skill-no-README- and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Roblox Agent Skill No Readme does

Drop-in skill that lets any AI coding agent (Claude Code, Codex, Antigravity, Cursor, OpenCode, GPT, Gemini) build, test, screenshot and publish Roblox games: Studio-first, Lune pipeline, fast lane, battle-tested on Don't Wake the Pumpkin King. roblox, roblox-studio, luau, lune, ai-agents, claude-code, codex, cursor, antigravity, agents-md, game-d

Alternatives in Testing

Develop Web Game — Build and test web games iteratively using Playwright with time-stepping 14.6k ★
Testing Strategy — This project is small, runs in a terminal, and is mostly deterministic 10.9k ★
Spec Kitty — Spec-Driven Development for serious software developers 1.6k ★

README

roblox-agent-skill

A portable, agent-agnostic **Roblox game development skill**: the rules, workflow, scripts and worked examples that shipped 144 versions of *Don't Wake the Pumpkin King* with AI agents — packaged so any agent (Claude Code, Codex CLI, Cursor, OpenCode, Google Antigravity, Gemini CLI, ChatGPT/Claude projects) and any frontier model can work on a Roblox game the same way.

What it gives an agent:

Standing rules that keep the game owner in control: everything visible is a real, editable Studio instance ("I WANNA BE ABLE TO CHANGE AND EDIT ANYTHING I WANT"), tunables are attributes, UI clones existing templates, "DON'T CHANGE THE DESIGN", the owner's game design is final, never claim a playtest you didn't run.
A fast loop: edit a build hook → tools/rig/go.sh builds the .rbxl with Lune (~2 s), renders PC + phone screenshots (~1 s), runs guard tests (~3 s) and publishes through Open Cloud only if they pass.
A gameplay lane with tests that must fail on the base build, full-suite comparison, and parallel workers + an independent verifier.
Lessons from real bugs (weak Instance-keyed tables, TextScaled without TextWrapped, Lune meshes that don't render, Studio edits wiped by pipeline publishes, …).

Layout

AGENTS.md            main instructions (persona, rules, loop, decision table, report format)  ← start here
SKILL.md             Agent-Skills style entry point (name/description frontmatter) → docs index
docs/                workflow, studio-first rule, Lune pipeline, testing, publishing, Studio MCP, UI, monetization,
                     security, launch checklist, lessons, state templates, thinking, subagents, model setup, resources
docs/examples/       5 worked sessions: request → plan → commands → output → report
adapters/            ready-to-copy files: claude-code/ codex/ antigravity/ cursor/ opencode/ chat-only/ + install.sh/.ps1
tools/rig/           go.sh shot.sh fast.sh run_all.sh compare_runs.sh publis