/gauntlet-loop
GAME skill — pure Gauntlet aim prompt; build a game vs a named reference until the human stops it
$ npx -y skills add duolahypercho/gauntlet-loop --agent claude-codeHow it fires
How this command gets triggered: by you, by Claude, or both.
- Fires itselfClaude auto-loads it when your prompt matches the work.
- You can call itInvoke it directly when you want it.
- Slash command
/gauntlet-loop
Context preview
What this command does when you run it.
GAME skill — pure Gauntlet aim prompt; build a game vs a named reference until the human stops it
Command definition
gauntlet-loop.mddescription: GAME skill — pure Gauntlet aim prompt; build a game vs a named reference until the human stops it argument-hint: "[game to build] [optional: against REFERENCE] [optional: in STACK]"
Invoke **gauntlet-loop**. This is a **game** skill. Pure prompt only — no harness, no state machine, no helper scripts.
Read `~/.claude/skills/gauntlet-loop/SKILL.md` → `AGENTS.md` → `CLAUDE.md`:
1. Infer game nouns from the args (thing / reference game / stack). One question max if THING is missing. 2. Fill the three-paragraph prompt in AGENTS.md (`/loop` + ultracode). 3. **Run it** on the **game**. Fan out. Separate harsh critic. Blind compare to the real reference game. 4. Art gaps: image gen for sprites/textures; Blender MCP for 3D meshes — see `ASSETS.md` / `BLENDER_MCP.md`. Always import into the playable build. 5. Keep going until the human stops you. Never ask "continue?" 6. Do not invent capture tools, contracts, scoreboards, or round ledgers.
Task: $ARGUMENTS
GAME skill. Matt Shumer's aim prompt as a slash command — build a game against a named reference. Fill the three paragraphs. Run them. Stop when you stop it. No harness. No state machine. No helper scripts. Pure prompt.