Skip to main content
Skills are purpose-built workflows that help your Agent produce higher-quality investment analysis than a general chat answer. Where a general model writes a fresh essay every time you ask, a Skill follows a defined process, so the analysis stays consistent, repeatable, and auditable.

​
What a Skill defines

A Skill usually specifies:
  • The analytical framework — the lens the analysis uses
  • Data sources to inspect — which of Driven’s 245 data endpoints to pull, and which fields
  • Steps to follow — the order of operations
  • Output shape — how the result is structured
  • Quality checks — cross-verification before the answer is presented
Because the framework and data sources are fixed in advance, two people asking the same Skill the same question get comparably structured answers. That consistency is what separates a research workflow from a chat reply.

​
Two ways to invoke a Skill

Automatic matching. By default, Driven reads your prompt and selects the right Skill for the job. Ask to value a company and it routes to the valuation workflow; ask to find stocks meeting criteria and it routes to the screener. You do not have to know the Skill names. Manual selection. You can also invoke a Skill explicitly through the menu or the command line. This is useful when you know exactly which workflow you want, or when you want to override the automatic choice. Type the command or pick from the menu and the Agent runs that specific Skill.

​
Built-in Skills

​
How to ask for a Skill

You can ask naturally and let Driven route: 
Use Deep Research to analyze MSFT's AI growth story and risks.
Or describe the task without naming a Skill: 
Screen profitable US software companies with strong free cash flow and reasonable valuation.
Both work. The first names the Skill directly; the second lets automatic matching pick it.

​
Quality tip

Ask the Agent to show its assumptions, evidence, and any missing data. This makes any Skill’s output easier to audit: 
Show the assumptions, the evidence behind each, and flag any data you could not retrieve.