I'd be interested in how this could be used. The $100 cap is the right shape of mitigation in terms of guardrails. Buying a domain is irreversible but has a price tag, so you can deterministically bound how irreversible it gets.
LLM based probabilistic systems are good (or bad in this case) at deciding what to do, and deterministic systems are good at carrying it out. Your deployment system should always be deterministic.
What I like about this is the inclusion of the Claude code pieces. I know there is a lot out there, but seeing a project that uses AI to produce features of this standard is great to see. I'd be interested in learning how the author developed features while building this out and how effective Claude/AI was while performing this building. Heck, I'd watch something where the author demonstrates their workflow. Nice job.
This hits close to home. I'm UK-based remote working for a US company, and I've seen this play out more times than I'd like. Led the architecture and design on products that went on to do $100m+, only for someone else to waltz in and take the credit once I'd moved on to the next thing.
The annoying bit is that being good at your job often means you get dragged into the next hard problem before the last one's had its moment. Meanwhile, whoever happens to be standing there when the champagne corks pop gets all the credit.
The paper trail advice is bang on. I'd just add - document decisions as you make them, not after the fact. Architecture decision records, design docs with your name on them, commit histories that tell the story. Handy when people's memories start getting conveniently fuzzy.