Loopit is a tiny convenience layer on top of Superpowers for Codex and Claude Code. It adds two slash commands for executing implementation plans without retyping the workflow every time.
Unofficial project: Loopit is not affiliated with Superpowers, OpenAI, or Anthropic. It is just a thin wrapper around existing Superpowers workflows.
- Codex or Claude Code
- Superpowers — if it is not installed yet, the skills will guide setup for your environment.
Use when you want controlled, chunk-by-chunk execution.
Use it instead of "implement this" when you want the agent to pause after each meaningful step, show what changed, and let you steer before it continues.
approved plan
-> implement next chunk
-> verify
-> short report
-> wait for "ok"
-> repeat until done
-> final verification
Use when the plan is clear and you want to hand execution over to the agent.
Use it when the plan is ready and you do not want to sit next to the agent, but still want it to follow a proper engineering workflow.
approved plan
-> Superpowers subagent-driven development
-> per-task review/fix loop
-> final review
-> final verification
-> branch finish / PR workflow
/loopit docs/superpowers/plans/my-feature.md runs that plan chunk by chunk and waits for approval after each chunk.
/loopitauto docs/superpowers/plans/my-feature.md hands that plan to the autonomous Superpowers execution flow.
/loopit uses the approved plan already present in the conversation and runs it with human checkpoints.
/loopitauto uses the approved plan already present in the conversation and runs it autonomously.
Clone the repo, then run the helper for your environment:
# Codex
scripts/install-codex.sh
# Claude Code
scripts/install-claude.shThe helpers symlink skills/loopit and skills/loopitauto into your agent's skills directory. Restart the agent after installing.
If Superpowers is missing, the skills stop before implementation and guide setup for the current environment.
Remove the symlinks and restart the agent:
rm -f ~/.codex/skills/loopit ~/.codex/skills/loopitauto
rm -f ~/.claude/skills/loopit ~/.claude/skills/loopitautoKeep Loopit a thin wrapper over Superpowers. Prefer small improvements that clarify routing, installation, or safety without copying Superpowers workflows into this repo.
MIT