tools I use for vibe coding.

This repository contains tools designed to help Claude Code (and other AI assistants) delegate work to background worker processes instead of attempting complex fixes directly in the main conversation context. This approach prevents context corruption and improves overall task completion quality.

I created this because Claude Opus and Sonnet (4) can be really bad sometimes. So o3helper let's them call o3. Right now it doesn't do patches very well so I mostly have them call for advice or to explain why something they implemented is bad.

This is a semi-agentic human in the loop planning system. Tell the LLM what you want then tell it to use @PLAN.md to create a plan in wherever/nameofplan-p1/. Then in a clean context do each "task" individually and use the "a" version of the task to verify. This produces better results on average than not doing this.

Documentation for launching background worker processes to handle complex tasks without corrupting the main AI assistant's context. This allows Claude Code to delegate work effectively rather than attempting everything in a single conversation thread.