--dangerously-skip-permissions in CI: where AI agent YOLO flags turn prompt injection into full access

Here is a GitHub Actions workflow that is easy to write on a Friday afternoon. It is a made-up example, but each piece of it is copied from patterns I keep seeing in public repositories: name: AI tria