Browse the specific Artifact Definitions Cuddler publishes for concrete artifact types, and use the role-based AI workflow to orchestrate a synchronized Data definition plus Template and Process documents that consume the published standards.
Build A New Artifact Definition With AI
Start by selecting the Document Role that matches the artifact's primary job. Then give an AI assistant the matching Document Role Specification, the shared Artifact Specification, a local orchestration payload, and the example documents you want it to follow. The goal is a synchronized local bundle: create or refresh the Data definition, one sanitized example Data Instance, and the concrete Template and Process documents together, keep them aligned through `.definition.json` inside `outputPath`, embed the parsed JSON copied from `sourceDefinitionPath` into that definition file, and follow the governing `for-ai` guidance from the published standards. Only the Data member is locally authored as a schema. The Template and Process members must consume the published Template and Process schemas from `www.cuddler.dev` instead of inventing family-specific local schemas.
Exact workflow
Choose whether the artifact is mainly Data, Template, or Process.
Open that role's Document Role Specification and the shared Artifact Specification.
Prepare a JSON payload with `outputPath` and `sourceDefinitionPath`; point `sourceDefinitionPath` at the source-side JSON context file you want embedded, and the Process side resolves automatically to the single `*.process.json` contract.
Keep your example documents in the same folder tree as `sourceDefinitionPath` so the assistant can discover them locally without widening the standards.
Have the assistant resolve precedence in this order: Artifact Specification, active role specification, `sourceDefinitionPath`, then the source-definition-folder example corpus discovered by scanning that folder tree and excluding `sourceDefinitionPath` itself.
Have the assistant create or refresh `.data.schema.json`, `.data-example.json`, `.template.json`, and `.process.json` inside `outputPath`, then write `.definition.json` there so `selectedDefinitions` points at the current bundle using definition-file-relative paths such as `.data.schema.json`, and `inputs.sourceDefinitionContent` stores the parsed JSON copied from `sourceDefinitionPath`.
Require the assistant to author node-level `for-ai` Alpaca arrays where the Data definition or Process questions need them and to consume existing governing `for-ai` guidance when drafting the concrete Template output.
Require the assistant to validate the generated role output against the correct cached normative target, and when Template is in scope require schema validation plus deterministic template semantic checks. When Process is in scope require one `*.process.json` document whose typed questions use explicit answer types, `conversation` stored as a JSON array of JSONL-style record objects, question-level `for-ai`, deterministic routing, and completed-conversation-to-output evaluation guidance.
If examples conflict or a business fact is still missing, have the assistant record the unresolved assumption briefly outside the produced file instead of broadening the contract.
What to include in the prompt
The chosen Document Role Specification JSON URL for Data, Template, or Process.
The shared Artifact Specification JSON URL.
A JSON payload that defines `outputPath` and `sourceDefinitionPath`, using absolute local paths and pointing `sourceDefinitionPath` at the source-side JSON context file you want embedded.
The example documents you will attach or paste into the conversation.
A request to convert every knowable property into definitive schema constraints rather than optional or ambiguous fields.
An instruction to keep the synchronized bundle aligned, to keep `.data-example.json` sanitized and free of PII, and to update current files in place rather than inventing version-suffixed variants.
A request to produce the canonical slug, exact schema-file inventory, exact validation responsibility for each published file, and the current-file paths selected by the definition file.
A direction to treat missing or conflicting examples as brief unresolved assumptions outside the schema rather than as permission to widen optionality.
A direction to list only unresolved assumptions instead of inventing a local workflow wrapper file.
Use the launch buttons below if you want ChatGPT preloaded with the right role-specific instructions.
Data role
Data Artifact Definition Prompt
Choose the Data role when the artifact's main job is to store, validate, exchange, or preserve structured facts.
The first JSON link is role-specific and points to /standards/document-role/data/v1.0.0/document-role-specification.json. Only the Artifact Specification stays shared across all three role cards.
What the assistant should extract
From the Data Role Specification, extract the artifact-purpose summary, the required document envelope, the stable metadata and identifier rules, and the domain validation expectations a Data-governed Artifact Definition must enforce.
From the Artifact Specification, extract the shared authoring metadata, governing-document links, source URL, attribution notice, confidentiality rule, and publication-bundle expectations that every Artifact Definition must publish.
Treat this as one member of a synchronized local artifact bundle and keep its artifact identity aligned with the companion example Data Instance, Template document, and Process document through the definition file.
Turn every knowable constraint into schema logic: fix language, date and date-time handling, meta.status, exact table or section counts, exact published schema files, and structural rules with const, enum, pattern, format, minItems, maxItems, required, and closed-object constraints instead of leaving them ambiguous.
Use the user-provided example documents only to derive business-domain facts, labels, and content expectations that cannot already be fixed by the governing standards, and do not infer optionality from missing or conflicting examples.
Example documents you should provide
Attach representative Data Artifact Documents or source materials that show the facts, sections, and level of specificity the new artifact type must support.
The first JSON link is role-specific and points to /standards/document-role/template/v1.0.0/document-role-specification.json. Only the Artifact Specification stays shared across all three role cards.
What the assistant should extract
From the Template Role Specification, extract the template-document envelope, markdown AST surface, renderer option rules, output metadata, semantic validation rules, and sanitization constraints that the local <artifactId>.template.json document inside outputPath must honor.
From the Artifact Specification, extract the shared Artifact Definition metadata, governing-document links, source URL, attribution notice, confidentiality rule, and publication-bundle expectations.
Treat this as one member of a synchronized local artifact bundle and keep its artifact identity aligned with the companion Data definition, example Data Instance, and Process document through the definition file.
Turn every knowable report rule into definitive document structure: fix language, date and date-time handling, renderer options, exact section or markdown-block counts, exact headings, exact table headers, exact reference identifiers, and markdown structure rules directly in the authored template document instead of leaving them ambiguous.
Use the user-provided example documents only to derive business-domain report content, labels, and reader-visible expectations that cannot already be fixed by the governing standards, and do not infer optionality from missing or conflicting examples.
Example documents you should provide
Attach representative rendered outputs or authored markdown-template materials that show the template sections, table shapes, narrative tone, and delivery expectations the new artifact type must support.
* For best results, configure ChatGPT to run in 'Thinking' mode.
Process role
Process Artifact Definition Prompt
Choose the Process role when the artifact's main job is to guide workflow behavior, automation, escalation, or execution order, especially for SOPs and business processes with branching or approvals.
The first JSON link is role-specific and points to /standards/document-role/process/v1.0.0/document-role-specification.json. Only the Artifact Specification stays shared across all three role cards.
What the assistant should extract
From the Process Role Specification, extract the typed question contract, identifier and path rules, conditional section rules, semantic validation rules, and published questionnaire boundaries that the local <artifactId>.process.json document inside outputPath must honor.
From the Artifact Specification, extract the shared Artifact Definition metadata, governing-document links, source URL, attribution notice, confidentiality rule, and publication-bundle expectations.
Treat this as one member of a synchronized local artifact bundle and keep its artifact identity aligned with the companion Data definition, example Data Instance, and Template document through the definition file.
Turn every knowable process rule into definitive document structure: fix language, date and date-time handling, meta.status, exact question counts when they are knowable, identifier formats, ordering rules, and conditional routing or section gates directly in the authored process document instead of leaving them ambiguous.
Use the user-provided example documents only to derive business-domain workflow content, actor labels, decision criteria, question wording, and operational details that cannot already be fixed by the governing standards, and do not infer optionality from missing or conflicting examples.
Example documents you should provide
Attach representative process source materials such as SOPs, business process maps, workflow documents, prompts, scripts, or equivalent operational artifacts that show the branching, approvals, exceptions, and workflow structure the new artifact type must encode.
* For best results, configure ChatGPT to run in 'Thinking' mode.
Each launcher uses the shared Artifact Specification plus the matching Document Role Specification and expects you to provide the local orchestration payload separately. The assistant should treat examples only as business-domain evidence, keep the synchronized bundle aligned through `.definition.json` inside `outputPath`, embed the parsed JSON copied from `sourceDefinitionPath`, keep `.data-example.json` sanitized and free of PII, and update current files in place instead of creating speculative version variants. Only the Data member is locally authored as a schema; the Template and Process members should consume the published schemas from `www.cuddler.dev`. For Process outputs, it should always author one `*.process.json` document and decide whether the question flow stays linear or branches through deterministic routing inside that same document.
