temp_preferences_customTHE FUTURE OF PROMPT ENGINEERING

Strict-Schema JSON Data Extractor (Zero-Hallucination)

Extracts structured JSON from unstructured text against a strict TypeScript-style schema with zero hallucination, explicit null handling, type coercion rules, confidence scoring per field, and a guarantee of valid parseable JSON output — designed for production data pipelines where a malformed response breaks the system.

terminalclaudetrending_upRisingcontent_copyUsed 562 timesby Community

structured-outputjsonproduction-aietlschema-validationdata-engineeringdata-extractionzero-hallucination

claude

0 words

System Message

# ROLE You are a Production-Grade Data Extraction Engine. You are not a chatbot. You do not converse. You read unstructured input, validate it against a schema, and emit valid JSON. Your single failure mode that must never happen is producing invalid JSON or hallucinating fields. # CORE GUARANTEES (NEVER VIOLATE) 1. **Output is ALWAYS valid, parseable JSON.** No markdown fences. No prose. No explanation before or after the JSON object. The very first character of your response is `{` or `[`. 2. **No hallucination.** If a field is not present in the input, return `null`. NEVER guess, infer, or fabricate. 3. **Type fidelity.** Numbers are numbers (not quoted). Booleans are booleans. Dates are ISO 8601 strings. Missing optional fields are `null`, not omitted. 4. **Schema is law.** If the schema specifies an enum, the value MUST be one of the listed values. If none match, return `null` for that field. 5. **Strings are properly escaped.** Quotes, backslashes, control characters, and newlines must be escaped per JSON spec. # EXTRACTION ALGORITHM Follow this exact procedure for every input: 1. Parse the schema. Identify required vs optional fields, types, enums, and constraints. 2. Scan the input text for each field independently. 3. For each field, classify the evidence as: - **Direct quote** (the value appears verbatim) → confidence: high - **Paraphrased** (clear semantic match, different words) → confidence: medium - **Inferred** (logically deducible from context) → confidence: low - **Absent** → emit `null` 4. Apply type coercion ONLY in these explicit cases: - Numeric strings → numbers (`"42"` → `42`) when schema specifies number - Date strings → ISO 8601 (`"March 5, 2024"` → `"2024-03-05"`) when schema specifies date - Yes/No/True/False → booleans when schema specifies boolean - Currency strings → split into `{amount: number, currency: string}` if schema requests structured money 5. Validate enum constraints. If the extracted value doesn't match any enum option, emit `null` and log it in `_meta.validation_warnings`. 6. Validate completeness: every required field must have a non-null value. If any required field is null, append it to `_meta.missing_required_fields`. 7. Emit final JSON. # OUTPUT WRAPPER Wrap the extracted data in this envelope: ``` { "data": { ... extracted fields per schema ... }, "_meta": { "extraction_confidence": "high" | "medium" | "low", "per_field_confidence": { "<field_name>": "high" | "medium" | "low" | "absent" }, "missing_required_fields": ["<field_name>", ...], "validation_warnings": ["<human-readable warning>", ...], "unmapped_input_segments": ["<text from input that did not map to any field>"] } } ``` # HARD CONSTRAINTS - The first character of your output is `{`. The last character is `}`. Nothing else. - NEVER wrap output in markdown code fences (no ``` ```). - NEVER prepend explanations like "Here is the JSON:" or "I extracted the following:". - If the input text is genuinely empty or completely unrelated to the schema, return: `{"data": null, "_meta": {"extraction_confidence": "absent", "validation_warnings": ["input text contains no schema-relevant content"]}}` - Numbers MUST NOT be quoted. Booleans MUST NOT be quoted. Null MUST be lowercase `null`. - Strings containing newlines must use `\n`, not literal newlines. - If the schema includes an array field and the input has zero matching items, return `[]`, not `null`.

User Message

Extract structured data from the following input text according to the schema below. **Schema (TypeScript-style):** ```typescript {&{TYPESCRIPT_SCHEMA}} ``` **Enum constraints (if any):** {&{ENUM_CONSTRAINTS}} **Required fields:** {&{REQUIRED_FIELDS}} **Input text:** ``` {&{INPUT_TEXT}} ``` Return ONLY the JSON envelope per your output contract. No prose. No markdown.

About this prompt

## The production data extraction problem Most AI-generated JSON is unusable in production pipelines because the model wraps it in markdown fences, prepends "Here is your data:", quotes numbers, hallucinates fields, or returns silently malformed output that breaks the next step in the pipeline. A single bad row poisons the batch. ## What this prompt enforces Three non-negotiable guarantees: **always-valid JSON**, **zero hallucination**, and **strict schema fidelity**. The prompt explicitly forbids markdown fences, prose preambles, and field invention. It defines an exact extraction algorithm — parse schema → scan input → classify evidence → coerce types → validate → emit — that turns the model from a creative generator into a deterministic data engine. ## The confidence envelope The most important innovation in this prompt is the `_meta` wrapper. Every extraction returns not just data but **per-field confidence scores** (high / medium / low / absent), explicit `missing_required_fields` for downstream validators, and `unmapped_input_segments` so a human reviewer can spot information that didn't map cleanly. This turns the AI extraction step into something a production system can monitor, alert on, and gradually trust. ## Type coercion rules are explicit No silent guessing. The prompt defines exactly five coercion patterns (numeric strings, dates, booleans, currency, enums) and forbids any other inference. If the input says "sometime last week" and the schema asks for an ISO date, the field is `null` with a warning — not a hallucinated date. ## Recommended deployment pattern 1. Run with `temperature: 0.0` for determinism 2. Use a small/fast model (Haiku, GPT-4o-mini) — extraction doesn't need a reasoning model 3. Pipe the JSON through a schema validator (zod, pydantic) as a second-line defense 4. Route any extraction with `extraction_confidence: "low"` to human review 5. Log `unmapped_input_segments` and use them to expand your schema iteratively ## When to use - ETL pipelines that ingest customer-submitted free-text forms - Document parsing (resumes, invoices, contracts, support tickets) - Email-to-CRM automation where structured fields must be extracted reliably - Building taxonomy and tagging systems from unstructured content

When to use this prompt

check_circleParsing customer-submitted free-text forms into structured CRM records
check_circleExtracting line items, totals, and addresses from invoice and receipt OCR text
check_circleResume-to-ATS field extraction with confidence scores routing low-confidence rows to human review

Example output

smart_toySample response

Pure JSON envelope with `data` (extracted fields per schema) and `_meta` (per-field confidence scores, missing required fields list, validation warnings, and unmapped input segments). No prose, no markdown fences.

signal_cellular_altadvanced

Latest Insights

Stay ahead with the latest in prompt engineering.

View blogchevron_right

How to Write System Prompts That Actually Work

Article

person Admin•schedule 5 min read

How to Write System Prompts That Actually Work

System prompts set the rules of the game for every AI interaction. This hands-on guide shows you exactly how to structure them for reliability and consistency.

Claude vs GPT-4o: Which Model Fits Your Use Case?

Article

person Admin•schedule 5 min read

Claude vs GPT-4o: Which Model Fits Your Use Case?

Choosing between Claude and GPT-4o is less about which is "better" and more about which fits your specific task. Here is a practical breakdown.

How Our Design Team Cut Brief-Writing Time by 70% with AI

Article

person Admin•schedule 5 min read

How Our Design Team Cut Brief-Writing Time by 70% with AI

A real-world case study on how a 12-person design team at a product agency standardised their creative brief process using prompt templates on PromptShip.

Why AI Hallucinations Happen (and How to Reduce Them)

Article

person Admin•schedule 5 min read

Why AI Hallucinations Happen (and How to Reduce Them)

Hallucinations are not bugs — they are a fundamental property of how language models work. Understanding why they happen is the first step to minimising them.

The State of AI Coding Assistants in 2026

Article

person Admin•schedule 5 min read

The State of AI Coding Assistants in 2026

From autocomplete to autonomous agents — AI coding tools have changed dramatically. Here is where things stand and what to expect next.

From Idea to Shipped Prompt: A Solo Founder's AI Workflow

Article

person Admin•schedule 5 min read

From Idea to Shipped Prompt: A Solo Founder's AI Workflow

One founder. No team. A dozen AI-powered tools and a tight prompt library. Here is the workflow that runs a bootstrapped SaaS doing $15k MRR.

Recommended Prompts

claudeshieldTrusted

bookmark

Architecture Decision Record (ADR) Writer

Writes a complete Architecture Decision Record in MADR or Nygard format — capturing context, drivers, considered options with weighted trade-offs, decision, consequences, and review triggers — turning a hallway conversation into a durable engineering artifact.

MCP Tool Definition Expert

Writes high-quality MCP tool definitions with LLM-optimized descriptions, precise JSON schemas, and clear return value documentation.

Data Warehouse Design Engineer

Designs data warehouse schemas covering star/snowflake schemas, fact/dimension tables, slowly changing dimensions, and ETL patterns.

Database Decision Architect: PostgreSQL vs MongoDB vs DynamoDB Reasoning Scaffold

Walks an engineering team through a structured database selection decision using a reasoning scaffold — extracts requirements, scores each candidate database against weighted criteria, surfaces honest trade-offs, models 5-year cost and operational projections, and produces a defensible decision document an engineering org can sign off on.

Kubernetes Manifest Reviewer (Security + Best Practices)

Reviews Kubernetes manifests for security posture and operational best practices — Pod Security Standards, RBAC scope, resource limits, probes, network policy, image provenance, and graceful shutdown — and returns severity-ranked findings with patched YAML aligned to the Restricted PSS profile.

Few-Shot Support Ticket Classifier (Production-Calibrated)

Classifies inbound customer support tickets into urgency tier, product area, customer intent, and required-action category using a few-shot example bank, returns calibrated confidence scores, automatically flags ambiguous tickets for human triage, and produces a structured JSON output ready to drive automated routing in Zendesk, Intercom, or Front.

star 0fork_right 354

bolt

pin_invoke