Chatting with LLMs

Clean Markdown Output

By default, some assistants insert numeric references (like [1], [2]) in the body text to cite sources, or generate unwanted introductory and concluding text. In Markdown files that clutters the reading — the page’s “mechanics” should be invisible so the reader stays in the flow. You can override this behavior per file with explicit instructions.

Avoid Chatty Text

To prevent the AI from producing friendly chat like “Here is your code…” or “Hope this helps!”, force a strict response:

Apenas responda, não produza nada:

Prompts for clean text (no references)

Generate the Markdown content without including the numeric references in brackets.

Keep the file text clean, with no source citations in the body of the document.

Protecting nested code blocks

When asking for a Markdown file that contains code blocks, ask to wrap the whole Markdown in four backticks — that way the inner blocks (three backticks) don’t break rendering:

Wrap the entire Markdown in four backticks to protect the inner code blocks.

⚠️ The “no references” instruction applies only to the file content; in a normal chat reply, citations are still useful for accuracy and transparency.