Built with Claude Code in a weekend

I subscribe to Claude Code, so I used it to build ZParen from scratch.

I used pstack and told Claude Code to "create an IDE to edit, debug and manage Common Lisp projects". That is a shorter version of what I wrote.

After some experiments, I also chose Rust and Tauri for the app. Some people may ask why I did not write all of it in Lisp. In my opinion, it is better to use the right tool for each job.

Skills and agents involved #

With pstack's /pstack:poteto-mode, Claude Code kept working while I slept. It loaded these skills while it built ZParen:

Skill	What it was used for
`pstack:poteto-mode`	The working mode: playbooks for features and bug fixes, and the principles below
`figure-it-out`	Planning the large multi-phase build-out
`feature` playbook	Each new feature, starting from its data shape
`show-me-your-work`	The decision trail kept during long autonomous runs
`technical-writing`	This README, the docs and the commit messages
`no-comments`	Keeping only comments that explain a non-obvious why
`claude-api`	The built-in assistant's Anthropic API client, model and streaming
Principles	`prove-it-works`, `foundational-thinking`, `model-the-domain`, `experience-first`, `exhaust-the-design-space`, `sequence-verifiable-units`, `laziness-protocol`, `build-the-lever`, `boundary-discipline`, `test-behavior-not-implementation`, `never-block-on-the-human`

Subagents took on separate pieces. Each coding agent worked in its own git worktree:

Agent	Work
`pstack:poteto-agent`	The SLIME-compatible indentation engine, paredit-style structural editing, the parinfer wasm crate, and the MCP server with its shared tool registry
`pstack:comment-sicko`	A pass that removed comments restating the code
`Explore`	Research into the UX of other Lisp IDEs
`general-purpose`	Side tasks, including an independent review of the decision trail