Gauntlet Loop

Run iterative build-critique loops that push AI output past "good enough" toward a concrete quality bar. Give a lead agent a goal and a real reference to compare against. The agent decomposes the goal into the smallest independently-improvable pieces, assigns each piece a builder and a separate critic with fresh context, and loops until the output matches or exceeds the bar. Use when the first draft is never good enough — games, websites, product design, writing, marketing campaigns, code, research, or anything where output can be inspected and improved.

9 stars 1 forksAI, LLMs & Agents

Install

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add sidneyswift/skills --skill gauntlet-loop
Manual install
git clone --depth 1 https://github.com/sidneyswift/skills.git /tmp/skills
mkdir -p ~/.claude/skills
cp -r /tmp/skills/gauntlet-loop ~/.claude/skills/gauntlet-loop

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# Gauntlet Loop

## Overview

Most agent workflows produce one result and stop. The result is
usually "pretty good for AI" — and that's where the bar stays.

A Gauntlet Loop raises the bar by adding a concrete reference and an
independent critic. The lead agent breaks the goal into the smallest
pieces that can be improved separately. Each piece gets its own
builder and a separate critic with fresh context. The critic compares
the output against the reference. If the reference wins, the critic
identifies the biggest remaining gap and sends the work back. The
loop continues until the output reaches the bar — or you decide it's
ready.

The method is simple: split, build, judge, repeat. The leverage comes
from the bar being real, the critic being independent, and the loops
running longer than you'd expect.

## When to use

- "Make this actually good, not just AI-good"
- "I want this to compete with the best examples in the category"
- "Keep improving it until it's great"
- "Build something ambitious — a game, a full site, a polished app"
- "The first draft is never good enough"
- Any task where the output can be inspected and iteratively improved
- Any task where you have (or can find) a concrete quality reference

## Core principles

### 1. Use an actual agent

Do not paste a Gauntlet Loop into a normal chat. Run it inside an
agentic harness — Claude Code, Codex, OpenClaw, or any environment
where the model can open files, run code, render results, inspect
Read the full skill on GitHub

Details

Skill name
gauntlet-loop
Path
gauntlet-loop/SKILL.md
Repo updated
9 November 2026

FAQ

What does the Gauntlet Loop skill do?

Run iterative build-critique loops that push AI output past "good enough" toward a concrete quality bar. Give a lead agent a goal and a real reference to compare against. The agent decomposes the goal into the smallest independently-improvable pieces, assigns each piece a builder and a separate critic with fresh context, and loops until the output matches or exceeds the bar. Use when the first draft is never good enough — games, websites, product design, writing, marketing campaigns, code, research, or anything where output can be inspected and improved.

How do I install the Gauntlet Loop skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add sidneyswift/skills --skill gauntlet-loop. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Gauntlet Loop popular?

It lives in sidneyswift/skills, which has 9 GitHub stars and 1 forks.

Related skills

Design Inspiration

9

sidneyswift/skills

Break through design blocks by systematically mining curated inspiration sources. When you're stuck on UI, brand, layout, or visual direction, this skill walks you through nine high-signal design galleries — each covering a different design surface — to find patterns, steal structure, and restart your creative momentum. Use when someone is stuck, needs visual references, wants to explore design directions, or asks "what should this look like?

Frontend & Design

Design Visual Experiences

9

sidneyswift/skills

Create or refine original digital visual experiences where art direction, motion, interaction, or spatial behavior is central. Use for expressive interfaces, interactive explainers, creative coding, and visual films; translate relevant references into implementation decisions and inspect the delivered result.

Writing & Content

Ecosystem Mapping

9

sidneyswift/skills

Take a core solution or transformation and map every friction point around it into a compounding product ecosystem — instead of building standalone products. Walks the user through identifying their core transformation, mapping friction points (before, during, after), spotting two-sided market opportunities, and designing products that strengthen each other. Use when someone has a working solution but is stuck on "what's next," wants to expand beyond a single product, or is thinking about leverage and ecosystem strategy.

Other

Gpt 56 Prompting

10

Archive228/lab-skills

Operating procedure for prompting the GPT-5.6 model family (gpt-5.6-sol/terra/luna) in agentic work: outcome-first prompts with explicit stopping conditions, autonomy and approval boundaries, tool routing, Programmatic Tool Calling, multi-agent (beta), reasoning effort and verbosity settings, and safe migration from GPT-5.5/5.4. Use whenever writing or revising system/developer prompts for GPT-5.6, wiring its tools, tuning reasoning.effort or text.verbosity, or debugging over-asking, over-stopping, or tool-loop behavior on these models.

AI, LLMs & Agents

Context Engineering

10

Archive228/lab-skills

Engineer an agent's context window as a finite attention budget: prompt at the right altitude, give judgment instead of rigid rules, use expressive tools and progressive disclosure over piled-up instructions, delete redundancy, and manage long-horizon runs with compaction, notes, and sub-agents. Trigger whenever you write or audit a system prompt, CLAUDE.md, skill, or tool set for an agent; when an agent ignores instructions or degrades on long runs; or when migrating prompts to Claude 5-generation models.

AI, LLMs & Agents

Graph Max

9

sidneyswift/skills

Turn any hand-drawn workflow graph into a working multi-agent script. Draw the graph in any tool (whiteboard, paper, Excalidraw, napkin), photograph or screenshot it, send it to a coding agent with a single prompt, and get a running orchestration script back. The skill teaches graph literacy — how to draw nodes, edges, fan-outs, feedback loops, and decision gates so the coding agent can implement them faithfully. Use when someone has a workflow in their head but no code, wants to orchestrate multiple agents, or asks "how do I wire these steps together?

AI, LLMs & Agents

Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.