CONSORT: Flexible ways for specifying the format of the output of a #Consort #task, #named #agent, or #pipeline

#CONSORT #Structured #English for #AI
Flexible ways for specifying the format of the output of a #Consort #task, #named #agent, or #pipeline:

1. Formal JSON Schema
! Extract structured user data from unstructured bio text
# Free-text bios pasted from a signup form, may be messy or incomplete
$ Return valid JSON only, no prose, no markdown fences
%252:
{
  “type”: “object”,
  “properties”: {
    “name”: {“type”: “string”},
    “email”: {“type”: “string”, “format”: “email”},
    “age”: {“type”: “number”},
    “tags”: {“type”: “array”, “items”: {“type”: “string”}}
  },
  “required”: [“name”, “email”]
}

2. Less formal JSON Template notation
! Extract structured user data from unstructured bio text
# Free-text bios pasted from a signup form, may be messy or incomplete
$ Return valid JSON only, no prose, no markdown fences
%84:
{
  “name”: “string”,
  “email”: “string”,
  “age”: “number”,
  “tags”: [“string”]
}

Only % directives are #framed here, because its JSON payload contains {, :, and other punctuation that a parser could otherwise misread — and the shorthand types (“string”, “number”) replace the JSON Schema version for brevity, at the cost of not being machine-validatable.

#CONSORT #Specification: https://hyperonomy.com/2026/08/03/consort-prompt-dsl-system-prompt/