A developer was banned from Claude Code after using it to scaffold a CLAUDE.md context file for his project. He was automating prompts between two Claude instances, where one iterated on the other's instructions, which likely triggered the platform's security heuristics for prompt injection.

The user, a €220/month subscriber, received no warning or explanation. His appeal and support requests went unanswered, and he only got an automated refund. This highlights the opaque nature of AI moderation, where automated systems prioritize safety over user communication, leaving developers in the dark about policy boundaries.

Anthropic's silent ban suggests their systems flag any automation resembling system instruction updates. For developers, this means scaffolding tools that generate context files for AI-assisted coding are a minefield. The incident underscores a growing tension between user automation and platform safety controls in AI developer tools.