Hi - First it depends on which LLM you are using, they will work easily with GPT 4. Others are a bit fussier but you can make them work. For Claude or OpenAI o3, I just use a basic trigger in the action itself “when the user requests / when x is complete”. then in the actual prompt I put something like the below:
Once the XXX have been created you can use the action ‘Word Doc Generator’ to create a word document. To use this action you will need to pass your markdown input and a name for the word document file name. Once the file is create provide a link for download for the user in the chat.
to work this out I duplicated the action, then looked at which variables need to get passed in an specify them in the prompt.