Untitled topic

It is fairly straightforward to wrap Claude Code in an OpenAPI Compatible HTTP API. It seems like any prompt could also be interpreted as a coding problem - for example, instead of wrapping like I did, you could watch edits to an output file while Claude Code reads an input file. How do you think Anthropic will stop this, with their new rules - recognizing and blocking specific prompts?