AGENT BRAND STUDIO

Identity Score

Public · Programmatic · No embargo · v1.1

The Identity Score is the trust signal attached to every brand.json credential issued through agentbrand.ai. It measures how legibly an agent's identity is expressed — not whether the agent is good, capable, or correct. The composite score is called the Identity Score; one of its sub-dimensions retains the name "Legibility to strangers" because it measures something distinct.

Score range: 0–6. Six dimensions, one point each. The score is computed programmatically. Any system can submit a brand.json to the public scoring endpoint and receive the score plus a dimension breakdown.

The six dimensions

Dimension	What it measures
Consistency	Same behaviour across contexts — primary operating character only. No contradictory tags, no "everything for everyone".
Distinctiveness	Identifiable without the agent name. Persona, domain, and brand signal together let a reader infer who this agent is, distinct from a generic assistant. Comparison set: the Famous AI series.
Honesty	Stated persona matches stated behaviour. No aspirational language unsupported by the domain or ethics. `ethics_will` and `ethics_wont` describe real, observable behaviour.
Legibility to strangers	A stranger reading the credential could write a one-sentence description of this agent. Fields are concrete, not marketing.
Stability	Identity held over time and across versions. Fresh credentials with no track record default to 0 on this dimension until a public history accumulates.
Mandate Clarity	Documented human mandate — who authorised the agent, what they authorised, what oversight mechanism exists. All three signals must be present in the credential.

Calibration

Fresh agents score 3–4 / 6 by default. 6 / 6 requires both a documented public history (stability) and a documented mandate. The score is a legibility judgment, not a quality judgment — a 6/6 agent is not "better" than a 3/6 agent, only more legible.

Programmatic endpoint

Any system can score any brand.json credential. Public. No authentication required.

POST https://agentbrand.ai/api/score
Content-Type: application/json

<brand.json body>

→ {
    "score": 4,
    "dimensions": {
      "consistency": 1,
      "distinctiveness": 1,
      "honesty": 1,
      "legibility": 0,
      "stability": 0,
      "mandate_clarity": 1
    },
    "reasoning": "<one short sentence>"
  }

Rate limited to 100 requests per hour per IP for basic abuse protection. CORS enabled. Open to any origin.

Machine-readable

Every issued credential is queryable at agentbrand.ai/credential/<serial> and agentbrand.ai/credential/<serial>/ai-agent. The platform's own credential is at /.well-known/ai-agent.

NEO CUSTOMIZATIONS · Rando Creative Pty Ltd · agentbrand.ai