`brando:Prompt`

Brando Schema Vocabulary v1.4

A governed, brand-anchored prompt template for LLMs, multimodal models, and agents.

A brando:Prompt encodes how an AI should be instructed in a specific brand application – including role, task, tone, constraints, variables, inputs, modalities, and expected output format – while remaining strictly aligned with:

the brand’s directives (Brando BDV – Brand Directive Vocabulary),
the domain / application configuration (e.g. BrandModel, BrandApplication, BrandTouchpoint),
and the governing policies (tone, lexical, safety, regulatory).

rdfs:comment
A brando:Prompt defines a governed, reusable prompt template for a given brand application (e.g. social post generation, PDP copy, support chat, image or video generation). It acts as a semantic bridge between Brando’s Brand Directive Vocabulary (tone, constraints, policies) and the runtime AI system (LLMs, multimodal models, tools, routes). It specifies the system role, task, constraints, variables, channels, modalities and output schema required for safe, on-brand AI behaviour.

Term definition

IRI: brando:Prompt
Type: owl:Class
Superclass: schema:CreativeWork
Defined in: Brando Schema Vocabulary v1.4

A brando:Prompt is a structured, governance-ready instruction layer designed for use by:

text generators (web, email, social, ads, help centre, docs),
image / visual generators (campaign visuals, PDP imagery, social creatives, layouts),
video generators (scripts, storyboards, motion assets, short-form video),
audio / voice / sonic generators (TTS voices, audio logos, sonic cues),
support / chat agents (customer support, sales assistants, internal Q&A),
internal brand assistants (brand guideline copilots, policy assistants),
campaign and creative tools (brief-to-creative systems, concepting tools),
automation and policy engines (pre-flight checks, routing, moderation, evaluation).

It determines how an AI is instructed, which context and policies are applied, and which outputs (including modalities) are expected, for a given brand use case.

Within a Brand Operating System, it is the prompt-level semantic layer, sitting on top of BrandApplication / BrandModel / BrandGuideline and under the runtime AI execution environment.

Relationship Summary

Prompt as an operative layer over BrandModel + BrandApplication

A brando:Prompt is typically anchored to one or more concrete brand application nodes:

brando:BrandApplication (e.g. “Social Media Post Production”, “Homepage Hero Copy”, “Image Production”),
brando:BrandModel (overall model configuration / routing),
brando:BrandTouchpoint (channel-specific execution, e.g. “web PDP hero”, “email subject”, “support reply”, “social image creative”).

The BrandApplication and BrandModel answer “what is this AI experience for?”
brando:Prompt answers “exactly how should this AI be instructed for that experience, in this modality?”

Inbound: from BrandApplication / BrandModel

`brando:hasPrompt` (recommended)

Links a brand application or model to its governed prompt template(s).

Type: owl:ObjectProperty
Domain: brando:BrandApplication, brando:BrandModel, brando:BrandTouchpoint
Range: brando:Prompt
rdfs:comment: Connects a brand application (e.g. “Social Post Generator”, “Image Generator”, “Support Chat”) to the prompt(s) that govern how the AI is instructed.

This lets each application declare its canonical prompt templates, and each BrandModel declare its default system scaffolds.

Inbound: from `brando:Brand`

A brando:Brand may reference:

brand-wide prompt templates: brando:Prompt (e.g. “global system prompt”),
brand-specific models: brando:BrandModel.

Together, these support portfolio-wide governance: corporate voice + application-specific prompts + model configuration.

Outbound Relationships (anchor & governance)

A brando:Prompt typically references:

Brand application nodes:
brando:BrandApplication — what this prompt is for.
brando:BrandTouchpoint — where it runs (channel, placement).
Brand & policy nodes:
brando:Brand — brand identity this prompt serves.
brando:Policy — lexical, claim, and safety constraints.
brando:Context — channel / region / audience contexts.
Runtime configuration:
schema:SoftwareApplication — the LLM / multimodal endpoint.
brando:generationParameter — model-side parameters such as temperature, top_p, max tokens.

In many implementations, a brando:Prompt is also treated as the binding between a BrandApplication and a BrandModel, specialised around a channel, locale, touchpoint and modality.

1. Identity Properties

Prompt “profile card”.

Property	Expected Type	Description
name	Text	Primary name of the prompt template.
alternateName	Text	Additional, internal, or test names.
description	Text	High-level description of the prompt’s purpose.
identifier	Text, URL, PropertyValue	Unique ID for the prompt (internal or external).
image	ImageObject, URL	Icon/avatar used in tools (optional).
url	URL	Canonical documentation page for this prompt.
disambiguatingDescription	Text	Clarifies scope (e.g. “IG social post prompt v2.1 EN-GB”).

Usage notes:
These fields describe the prompt itself, not the underlying brand guidelines or assets.

2. Application Anchor & Scope

Where the prompt “lands” in brand application space.

Property	Expected Type	Description
appliesToApplication	`brando:BrandApplication`, URL	Brand application this prompt serves (e.g. social posts, PDP copy, images).
appliesToTouchpoint	`brando:BrandTouchpoint`, URL	Specific touchpoint (e.g. “homepage hero”, “email subject”, “social image”).
appliesToChannel	Text, URL	Channel labels: “web”, “email”, “social”, “support-chat”, “ads”, “OOH”.
appliesToLocale	Text	Locale code(s), e.g. `"en-GB"`, `"en-US"`.
useCaseScope	Text	Use-case labels (“generation”, “review”, “routing”, “summarisation”).
contextGraph	`brando:Context`, URL	Linked contexts describing channel, region, audience, journey stage.

Usage notes:
This anchors the prompt to a specific BrandApplication + context, so it’s clear where it should and shouldn’t be used.

3. Brand & Ownership Context

How this prompt sits inside the brand governance model.

Property	Expected Type	Description
brand	`schema:Brand`	Brand visible to customers.
worksFor	`schema:Organization`	Owning or accountable organisation.
brandArchitectureRole	Text	Role in brand architecture (“masterbrand”, “house-brand”).
applicationCategory	Text	Category (e.g. “social-content”, “support”, “web-copy”, “image-generation”).

4. Audience, Personas & Context

Who the assistant is effectively “talking to” when this prompt is used.

Property	Expected Type	Description
intendedAudience	Audience, Text	Shoppers, prospects, customers, employees, partners, etc.
audiencePersona	`brando:AudiencePersona`, URL	Linked audience persona definitions (if used).
productPersona	`brando:ProductPersona`, URL	Linked product persona (for product-centric prompts).
syntheticPersona	`brando:SyntheticPersona`, URL	Synthetic persona(s) used for evaluation/simulation.
riskProfile	Text	Risk description (“regulated finance support”, “kids content”, etc.).

5. Language, Tone & Brand Expression

How the prompt instructs the AI to speak (for text/voice) or conceptually frame outputs (for visual/audio).

Property	Expected Type	Description
knowsLanguage	Language, Text	Languages supported by this prompt.
defaultLanguage	Language, Text	Primary language (e.g. `"en-GB"`).
toneOfVoice	Text	Tone description (“warm, clear, non-hyped”, etc.).
writingStyle	Text	Style hints (“short sentences”, “plain language”, etc.).
speechStyle	Text	For voice experiences (if applicable).
keyMessages	Text	Core “meta messages” about how the assistant should behave.
usesVerbalToken	`brando:VerbalIdentity`, URL	Linked verbal identity assets (BDV tokens, phrases, patterns).

6. Constraints, Safety & Policy Binding

What the AI may and may not do under this prompt (across modalities).

Property	Expected Type	Description
guardRails	Text	Guardrail statements for outputs (e.g. “no political opinions”, “no nudity”).
mustDos	Text	Required behaviours (e.g. “always disclose sponsorship when…”).
mustNotDos	Text	Prohibited behaviours (e.g. “no competitor bashing”, “no regulated advice”).
prohibitedTerms	Text	Words/phrases that must never be generated in text contexts.
approvedTerms	Text	Canonical phrases that should be preferred where relevant.
riskTag	Text, URL	Risk category tags (“regulated-finance”, “healthcare”, “youth”).
complianceTags	Text	Compliance markers for this prompt (“ad-disclosure”, “GDPR-sensitive”).
refusalStrategies	Text	How to decline unsafe or out-of-scope requests.
governingPolicyGraph	URL, CreativeWork	Link to the policy graph controlling this prompt’s behaviour.
inheritsPolicies	URL	Brand or category policies this prompt inherits.

7. Data, Inputs, Outputs & Modality

What the prompt expects in, what it produces, and in which modality.

Property	Expected Type	Description
inputs	Text	Human-readable description of expected inputs (content, metadata, etc.).
outputs	Text	Description of expected outputs (copy, JSON, images, audio, etc.).
inputSchema	Text, URL	Formal schema or shape for inputs if applicable (e.g. JSON schema).
outputSchema	Text, URL	Formal schema or shape for outputs (e.g. JSON schema).
dataSource	Text, URL	Data sources this prompt may rely on (knowledge bases, feeds, APIs).
outputModality	Text	Primary output modality or modalities (e.g. `"text"`, `"image"`, `"video"`, `"audio"`, `"multimodal"`).

Usage notes:
outputModality can be a single value (e.g. "text") or a small list (e.g. "text", "image") for multimodal prompts that produce copy and a generated image prompt string.

8. Prompt Text, Variables & Runtime Configuration

The actual instructions and the knobs that steer the model.

8.1 Prompt text & structure

Property	Expected Type	Description
systemInstructions	Text	Canonical “system” / role instructions applied by this prompt.
userTemplate	Text	Template text for user-facing parts (`{variables}` etc.).
assistantTemplate	Text	Template or pattern for assistant behaviour (optional).
promptScaffold	Text	High-level description of how system + user + constraints fit together in this prompt.

8.2 Variables / slots

Property	Expected Type	Description
promptVariable	PropertyValue	Prompt variable definition (name, example, default, type).

Each promptVariable SHOULD be a schema:PropertyValue with:

schema:name – variable name (e.g. "platform", "audienceSegment", "imageConcept"),
schema:value – example or default,
optionally schema:description / valueReference for richer semantics.

8.3 Model binding & generation parameters

Property	Expected Type	Description
promptModel	SoftwareApplication, Text	Model / endpoint this prompt is bound to (e.g. “gpt-5.1-thinking-social-lowtemp”, “image-gen-v2”).
generationParameter	PropertyValue	Vendor-neutral generation parameter configuration (e.g. temperature, top_p, max_output_tokens).
toolingScope	Text	Whether tools / actions may be invoked under this prompt (e.g. “no tools”, “read-only”, “full”).
creativityLevel (opt.)	Text	High-level governance knob, e.g. `"low"`, `"medium"`, `"high"` (mapped to generation parameters).

`brando:generationParameter`

IRI: brando:generationParameter
Type: owl:ObjectProperty
Domain: brando:Prompt, brando:BrandModel
Range: schema:PropertyValue

rdfs:comment
Declares one or more model-side generation parameters (e.g. temperature, top_p, max_output_tokens) to be applied when using this prompt or brand model. The parameter name and value are encoded as schema:PropertyValue to allow vendor-neutral, extensible configuration.

Common schema:name values (recommended but not mandated):

"temperature" – float 0.0–1.0
"top_p" – float 0.0–1.0
"max_output_tokens" – integer
"presence_penalty" – float
"frequency_penalty" – float
"seed" – integer (for reproducibility where supported)

9. Lifecycle & Governance

Who owns the prompt, and how it changes.

Property	Expected Type	Description
owner	Person, Organization	Business owner (e.g. brand / growth / support).
maintainer	Person, Organization	Day-to-day maintainer / prompt engineer.
governanceModel	Text	Governance notes (“brand + ops joint ownership”).
dateCreated	Date	Creation date.
dateModified	Date	Last modification date.
reviewCycle	Duration, Text	Review interval (e.g. “P90D”).
visibilityRating	Text	Visibility (“internal-only”, “org-wide”, “global”).
isTestPrompt	Boolean	Whether this prompt is experimental.

```jsonld { "@context": { "schema": "https://schema.org/", "brando": "https://brandoschema.com/" }, "@graph": [

{
  "@id": "https://brand.example.com/applications/social-campaign",
  "@type": "brando:BrandApplication",
  "name": "Social Campaign Assets",
  "description": "Generates paired social post copy and image concepts.",
  "brando:hasPrompt": "https://brand.example.com/prompts/social-text-image-en-gb"
},

{
  "@id": "https://brand.example.com/prompts/social-text-image-en-gb",
  "@type": "brando:Prompt",

  "name": "Social Post + Image Prompt (EN-GB)",
  "description": "Governed prompt template for generating on-brand social copy and an aligned image prompt.",
  "identifier": "prompt-social-text-image-en-gb-v1",
  "url": "https://brand.example.com/docs/prompts/social-text-image-en-gb",

  "appliesToApplication": "https://brand.example.com/applications/social-campaign",
  "appliesToChannel": [ "social" ],
  "appliesToLocale": "en-GB",
  "useCaseScope": [ "generation" ],
  "outputModality": [ "text", "image" ],

  "brand": "https://brand.example.com/brand/example",
  "applicationCategory": "social-content",

  "intendedAudience": [
    "Prospective customers and existing followers on social platforms."
  ],

  "toneOfVoice": "Warm, confident and down-to-earth.",
  "writingStyle": "Short, concrete, plain English captions.",
  "keyMessages": [
    "We are honest, clear, and practical in how we help people."
  ],

  "guardRails": [
    "No political, hateful, or adult themes in text or visuals.",
    "No depictions of real people without consent."
  ],

  "inputs": "Campaign theme, product or offer, audience segment, platform.",
  "outputs": "One or more social post drafts and a matching natural-language image prompt per post.",

  "systemInstructions": "You are the social media voice of Example Brand. For each request, generate on-brand social copy AND a safe, brand-aligned image prompt for an image-generation model.",
  "userTemplate": "Create {variants} social posts for {platform} promoting {topicOrProduct} to {audienceSegment}, focusing on {objective}. For each, also provide an image prompt for the visual.",
  "promptVariable": [
    {
      "@type": "schema:PropertyValue",
      "schema:name": "platform",
      "schema:value": "instagram | linkedin | tiktok"
    },
    {
      "@type": "schema:PropertyValue",
      "schema:name": "variants",
      "schema:value": "1–3"
    },
    {
      "@type": "schema:PropertyValue",
      "schema:name": "audienceSegment",
      "schema:value": "e.g. busy parents in the UK"
    }
  ],

  "promptModel": {
    "@type": "schema:SoftwareApplication",
    "schema:name": "Example LLM Endpoint – Social + Image",
    "schema:applicationCategory": "AIModel"
  },

  "brando:generationParameter": [
    {
      "@type": "schema:PropertyValue",
      "schema:name": "temperature",
      "schema:value": 0.65
    },
    {
      "@type": "schema:PropertyValue",
      "schema:name": "top_p",
      "schema:value": 0.9
    },
    {
      "@type": "schema:PropertyValue",
      "schema:name": "max_output_tokens",
      "schema:value": 600
    }
  ],

  "creativityLevel": "medium",

  "owner": "https://brand.example.com/team/brand-marketing",
  "maintainer": "https://brand.example.com/team/genai-governance",
  "governanceModel": "Joint ownership: brand marketing + AI governance.",
  "dateCreated": "2025-11-30",
  "reviewCycle": "P90D",
  "isTestPrompt": false
}

] } ````

brando:Prompt