Tips for good results
How to get more useful output from an installed persona.
Give it real context
Source documents, prior decisions, constraints, actual numbers. Personas are built to avoid inventing specifics, so the more grounded the input, the more useful the output. Thin input gets you a request for more detail rather than a confident-sounding guess.
Be specific about the deliverable
"Help me with requirements" is vaguer than "draft a BRD for this feature from these discovery notes." The more precisely you describe the artifact you want, the less back-and-forth it takes to get there, and the easier it is for Claude to route to the right persona if you have more than one installed.
Read the capability list first
Every persona's detail page in your library lists what it actually does. Skimming that before you start tells you whether you are asking for something in scope, and gives you the right vocabulary, the persona's own terms for its deliverables, to ask precisely.
Expect to be asked questions
A hard input gate means a persona interrupts with clarifying questions rather than filling gaps with invented content. That is by design, answering them fully up front is faster than iterating on an output built on guesses.
Review before you confirm
The confirmation gate exists so you see the actual artifact before anything irreversible happens. Use that pause, read what was drafted rather than approving on trust.