Create intermediate Works with: Any AI

Output Format Specification Writer

Structured brief Has example

Define exactly how the answer must look.

The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.

Write the output-format part of a prompt. The task: [TASK]. What will consume the output (a person, a spreadsheet, a program, a template): [CONSUMER]. Fields and rules I need: [FIELDS_AND_RULES].

Provide three levels of specification:
1. Plain structure for a person: headings in order, length per section, bullet or paragraph style, tone, and what to leave out.
2. A table specification: column names, what each contains, allowed values, and sort order.
3. A JSON specification: a schema with field names, types, allowed values, required versus optional, null handling, and a one-record example. Include the instruction "Return only valid JSON, with no commentary" where a program reads it.

For each, show a complete instruction block I can paste. Add rules for edge cases: what to output when information is missing (null or "not stated"), when there are multiple items, when the input is invalid, and when the request is outside the task. Include a validation checklist (does it parse? are all fields present? are the values in the allowed set?) and a short script outline to check JSON automatically. Warn that models can still produce invalid output, so validate it in code and retry on failure. Use my fields only; ask if something is unclear.

Fill in before use

Optional: type your details and the prompt above updates. Nothing is saved or sent.

Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.

For a program (JSON):
"Return only valid JSON, with no commentary: {"summary": string (max 40 words), "sentiment": "positive"|"neutral"|"negative", "action_items": [{"owner": string|null, "task": string}], "missing_info": [string]}. Use null when information is not stated."
For a person: headings in order (Summary, Risks, Next steps), 3 bullets each, plain tone.
Edge cases: empty input -> return the JSON with empty arrays; out-of-scope request -> {"error": "out_of_scope"}.
Validate output in code and retry on parse failure. Models can still return invalid JSON, so run a parser on every response, log failures, and keep the plain-text fallback for people who read the result. Example record: {"summary": "Supplier confirmed delivery for Friday.", "sentiment": "positive", "action_items": [], "missing_info": ["PO number"]}.

• output-format
• structured-brief