I am sure some variant of the infinite monkey theorem applies. Anthropic devs have unlimited access to Claude, so they are no doubt lazy when it comes to prompt writing and just brute-force the agent to do what they want.