A Living AWS "Gotcha" Doc for Coding Agents

Available as a Claude Skill

I've yet to find an LLM that can bang out sophisticated AWS apps from scratch. The fact is, AWS is great because it gives the builder so much control, but the trade-off is complexity.

Reference architectures can help if you know what you're building at the start. But this conflicts with the flow of modern agentic coding, where projects often begin with a fuzzy idea, then you gradually figure out what the product should do over several iterations. I call this "backing into architecture."

In practice, this means coding agents need to be able to look things up as they go. Agents like Claude Code are quite good at that, so I've been collecting AWS gotchas in a doc called AWS_LEARNINGS that I keep in GitHub. Since it's a living document, I made a Claude Skill called aws-architecture-gotchas that looks up the latest information whenever it needs it.

The great thing about Claude Skills is that I don't have to remember what each Skill is called. Claude just figures out if it's is needed from the context of the conversation and weaves it into the workflow.

If you want to contribute your own gotchas, let me know and we can figure out a way to crowdsource this.

You can install all of my skills by pasting this into Claude Code:
/plugin marketplace add jbdamask/john-claude-skills
/plugin install john-skills@john-claude-skills