Skip to main content
A skill is a named block of instructions you save once and reuse. If you find yourself typing the same framing into every Deep Dive question, that framing belongs in a skill. Typical skills:
  • A house definition of DSCR, DCOH or fixed costs, so every answer computes them the same way.
  • The output shape you always want, for example “return one HTML table, fiscal years as columns, round to USD thousands, use N/A for missing years”.
  • Sector rules, for example “for charter schools, exclude entrance fees from coverage”.

​
Writing a skill

Open Skills (/platform/skills) and click New Skill. Give it a name you’ll recognise in a dropdown, then write the instructions in the editor and click Save changes. Nothing saves until you do, and an Unsaved badge shows while there are pending edits. Discard changes puts the last saved text back. There is no length limit. The same advice applies as to any other prompt: see Writing effective prompts.

​
Using a skill in Deep Dive

In a Deep Dive session, type / in the question box. A list of your skills appears; keep typing to filter it, then pick one. The skill drops in as a labelled chip and you carry on writing around it. You can add more than one to a question. When you submit, the chip expands into the skill’s full instructions. The chip stays visible next to your question in the thread, so later you can see which skills that answer used.
A question pins the version of the skill that existed when you asked it. Editing or deleting a skill changes what future questions get; it never rewrites an answer you already have.

​
Sharing

Skills are private to you by default. Open a skill’s menu and click Share to make it visible to everyone in your organization, so the team starts from the same definitions. Shared skills appear for teammates under Shared with me and are read only: they can use one in a question but not edit it. Stop sharing takes it back. The same menu has Rename and Delete.

​
Skills and reusable instructions elsewhere

Skills are for Deep Dive. Data Tables and Tear Sheets carry their own saved instructions, since each of those is already built around one reusable prompt. When you create a new table or tear sheet you can load the instructions from an existing one as a starting point.