Skip to main content
If you’ve saved a prompt template on the Prompts page, you can run it against any entity from inside a Deep Dive session without copy-paste.

​
The /run-prompt flow

1

Open or create a session

Open a session for the obligor or CUSIP you want to analyse.
2

Scope the documents

Adjust the document filter bar (Years back, Disclosure types, Include latest OS) so the template runs against the right corpus.
3

Open the command menu

Type / in the chat box.
4

Pick a template

Select run-prompt and pick your template. A tag appears in the chat showing which template was run, so the history records exactly what was asked.
5

Review and iterate

Check the answer and its citations. If the scope wasn’t right, change the filters and re-run. If the prompt itself needs work, edit it on the Prompts page and re-run.

​
Why use a template instead of typing the prompt

  • Repeatability: the same template on two obligors produces comparable output, which is hard with ad-hoc questions.
  • Speed: a tuned prompt can be 500+ words; running it from the menu is one click.
  • Audit trail: the chat history records the template name, so reviewers know exactly what was asked.
  • Re-use in watchlists: the template you tuned here can be applied to every entity in a watchlist via an assignment.

​
Authoring and best practices

For everything about creating, editing, sharing, and writing better templates, see Prompt Templates. For how to phrase any prompt so it retrieves the right evidence, see Writing effective prompts.