LoopCoding.dev
design loops, don't prompt agents · a Claude Code skillGitHub↓ Download skill
A downloadable agent skill · no login, nothing to install online

Design loops.
Don't prompt agents.

LoopCoding is a skill you install into your coding agent. Run /loop-coding, answer a few questions, and it writes a self-verifying LOOP.md — wired to your real repo, skills, subagents, and tools — then hands you one command to run it.

↓ Download SKILL.mdView on GitHub
01 How to install
  1. Drop the skill into your agent's skills folder — pick your flavor:
    curl -fsSL https://loopcoding.dev/install.sh | bash
  2. Restart your agent so it picks up the new skill.
  3. It's now available as /loop-coding.
02 How to run a loop
  1. Run /loop-coding → answer a few questions → it writes LOOP.md.
  2. Start the loop with the command it gives you:
    /loop Follow ./LOOP.md — run the checks every iteration, respect every guardrail, and keep going until all DONE-WHEN boxes pass, then stop.
  3. The agent builds, tests, and self-corrects until every DONE-WHEN box passes.
You shouldn't be prompting coding agents anymore. You should be designing loops that prompt your agents.— Peter Steinberger (@steipete), Jun 2026
WHAT IS LOOPCODING