Use Claude Code to build from a spec with review agents

A commenter had Claude Code implement software from a written spec while sub-agents checked the code and tests ran throughout.

Done with Claude Code· Intermediate· Apr 2026

What they did
They put Claude Code in auto mode and gave it a detailed specification to follow. As it worked, code-review sub-agents looked over major pieces, context rules kept the session organized, and automated tests ran continuously. The human stayed in a supervisory role: let it run, check the results the next morning, decide on edits or refactors, update the spec, and then repeat.
How it went
They said the approach could be especially useful for solo developers and might bring big productivity gains. They did not share a specific project, final metric, or completed product, and they said the workflow only makes sense if the token cost is affordable.
Tips
This kind of setup works best when the spec is specific and you are willing to review each iteration. Budget for higher token use, and treat cost as part of the decision to use the workflow.

Try it yourself with Claude Code

Act as my coding partner for [project]. Use the spec in [spec file or pasted spec] as the source of truth. Work in small increments, keep your notes concise, run tests as you go, and use review sub-agents to inspect each major change before you continue. If something is unclear, ask me before guessing. Pause and ask for my approval before any irreversible action such as deleting files, submitting code, sending data, or making payments. When you finish a round, summarize what changed, what still needs review, and what you want me to decide next.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on reddit.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.