Back to docs

Tool Persona Create From Description

Category: Persona Creation Β· Tier: All

Create a persona from a text description. Returns a trait reference matrix for inferring values.

When to Use

Persona creation has three workflows: description (AI generates traits), questionnaire (structured interview), and manual trait entry. All produce full 25-trait cognitive profiles.

Parameters

Parameter Type Required Description
persona_name string Yes Name for the new persona
description string Yes Text description of the persona (e.g., 'An impatient power user who skims content')

Example

Use persona_create_from_description on your target page

Every trait must be accounted for

The response includes trait_criteria β€” what a description looks like for each of the 25 traits β€” and a completeness_contract. That contract is enforced: persona_create_submit_traits refuses a persona with unaccounted traits and writes nothing.

Each trait needs either a value inferred from the description, or a place in unsupportedTraits:

{
  "tool": "persona_create_submit_traits",
  "params": {
    "persona_name": "anxious-first-timer",
    "traits": { "patience": 0.8, "readingTendency": 0.9, "selfEfficacy": 0.35 },
    "unsupportedTraits": ["fearOfMissingOut", "anchoringBias", "socialProofSensitivity"]
  }
}

Both answers are valid. Naming a trait as unsupported is not a failure β€” a description that says nothing about someone's response to scarcity should not be made to produce a number.

What is not valid is silence. Any trait you leave out is filled from a population baseline, and the resulting persona looks complete from the outside: a full 25-trait vector, real numbers, nothing flagged. A defaulted trait and a considered one become indistinguishable the moment the response is discarded.

A rejection names the gaps and returns the criteria for exactly those traits:

{
  "error": "incomplete_persona",
  "unaccountedTraits": ["satisficing", "trustCalibration", "interruptRecovery", "..."],
  "traitCriteria": {
    "satisficing": "First adequate option vs the best one. Look for: settles, maximizer, seeks optimal, good enough."
  },
  "howToProceed": "Re-submit with a value for each trait the description supports, and the rest listed in unsupportedTraits. Nothing was written.",
  "written": false
}

Big Five, estimated from the same description

The response also returns behavioural anchors for the five personality factors. Supplying bigFive with bigFiveEvidence β€” a quote from the description per factor β€” moves the persona to a values route that reaches all thirteen motivational axes rather than the nine the cognitive-trait route can reach.

Every factor must quote the description. A profile of five exact 0.5s is rejected rather than stored: that is what "the description does not say" looks like wearing the costume of "this person is perfectly average".

Tips

  • Combine with other tools for full analysis

Related

From the Blog