Claude Code is amazing at writing code, but it will happily build from under-specified requirements, implement the same thing different ways, and fail to write great tests unless you specifically ask it to.

Prawduct is a set of prompts, skills, hooks, and artifact templates that help focus Claude Code on product development rather than code development.

You can start from something a simple as "make a website with a scientific calculator" or as complex as "create a MMO with clients for iOS, Android, and web". You can specify as much or as little arch standards or implementation details as you want.

Specialized skills like /critic and /janitor are run automatically and apply context-less reviews to catch drift, hacks, and violations of best practices.

I've been using Prawduct myself for a couple of months, developing my own projects and also iterating on Prawduct itself (which is of course self-hosted on its own framework).

I'd love to hear feedback.

Interesting. Where are you getting the hooks, prompts, and skills sourced from? Are these custom-built or pulled from an existing framework?

I think Prawduct addresses a real problem. Claude Code builds what you ask for, not what you need, and the drift/inconsistency over a session is where most quality breaks down. The automatic /critic and /janitor skills are the strongest idea here making review happen by default rather than on-demand is the right design IMO.