What two techniques are used to control LLM output to achieve a specific format, such as getting multiple commands without commentary?	Message prefilling and stop sequences.
What initial sequence was used to instruct the LLM to begin writing a code block?	Three backticks (```).
Why did the LLM initially add a language identifier (like "bash") when using three backticks?	It inferred the language identifier to apply syntax highlighting when rendering the markdown code block.
How can a user prevent the LLM from automatically adding an unwanted language identifier?	By explicitly including the desired language identifier in the prefilled message.
What is one potential issue that might still occur even after successfully generating a code block?	The LLM might insert explanatory comments (e.g., "This command does XYZ") inside the code block.
How can the message prefill be used to prevent the LLM from including explanatory comments?	By adding explicit instructions in the prefill, such as "here are all three commands in a single block without any comments."
What is the scope of message prefilling?	It is not limited to using specific characters (like backticks); it can be used to provide complex guidance to the LLM.