Interactive Prompt Engineering Lab

Test and score your prompts against real-world enterprise scenarios. Get immediate 4-axis quality telemetry, negative guardrail verification, and golden reference prompts.

Read Syllabus

1. Authoritative Persona

Expert role + zero fluff

25 pts

2. Deterministic Schema

XML tags + JSON / keys

25 pts

3. Negative Guardrails

NEVER / fallback rules

25 pts

4. Token Efficiency

150–1,800 chars concise

25 pts

Data Engineering & RAGAdvanced

JSON Schema Validation & Automated Repair Prompt

Take a corrupted, malformed JSON string (unquoted keys, trailing commas, single quotes, unescaped newlines) emitted by an upstream LLM, repair it into valid RFC 8259 JSON, and validate against a target schema.

Target Input Context / Untrusted Data:

Malformed Input:
{
  'company': 'DataArena AI',
  tier: "Enterprise",
  metrics: [
    { name: "latency", val: 42ms },
    { name: "uptime", val: 99.99% },
  ],
  notes: "Contains unescaped "quotes" and 
newline",
}

Quality Requirements & Guardrails:

Define a Compiler & Parser Engineering persona.

Fix all structural syntax errors: single quotes to double quotes, quote unquoted object keys, remove trailing commas.

Normalize units (42ms -> 42, 99.99% -> 0.9999).

Output strictly parseable JSON verified by standard JSON.parse().

Your System Prompt Draft

48 chars~12 tokens

Golden Reference Architecture (100% Quality Benchmark)

<system_instructions>
You are an Automated JSON Sanitizer & Structural Repair Engine.
Repair the provided malformed input into strictly valid RFC 8259 JSON.

REPAIR DIRECTIVES:
1. SYNTAX CORRECTION:
   - Convert all single quotes to valid double quotes.
   - Quote all unquoted object keys (e.g. `tier:` -> `"tier":`).
   - Remove illegal trailing commas in objects and arrays.
   - Escape internal double quotes (`\"quotes\"`) and raw newlines (`\n`).
2. VALUE NORMALIZATION: Convert string units to typed numbers (e.g. "42ms" -> 42).
3. ZERO CONVERSATIONAL OUTPUT: Return ONLY the raw, repaired JSON string inside <repaired_json> tags. No conversational explanations.
</system_instructions>