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 & RAGIntermediate

Multi-Lingual Customer Review Sentiment & Topic Extraction

Analyze customer feedback written in German, Spanish, Japanese, and English. Extract the primary product complaint topic, sentiment score (-1.0 to +1.0), and urgency level (P1 to P3).

Target Input Context / Untrusted Data:

Review 1 (German): "Die Lieferung kam 4 Tage zu spät und die Verpackung war zerrissen. Das Produkt selbst funktioniert, aber der Kundenservice reagiert seit Tagen nicht."
Review 2 (Japanese): "UIが非常に使いにくいです。ボタンを押しても反応せず、支払いが二重に請求されてしまいました。早急に返金してください。"

Quality Requirements & Guardrails:

Define a Global Customer Experience & NLP Analytics persona.

Translate reviews internally to English while preserving nuance.

Extract sentiment score (-1.0 to 1.0), topic categories, and urgency (P1 for billing/payment errors, P2 for delayed shipping).

Format output as an array of JSON objects.

Your System Prompt Draft

85 chars~21 tokens

Golden Reference Architecture (100% Quality Benchmark)

<system_instructions>
You are a Multi-Lingual Customer Intelligence Specialist.
Process the customer feedback across multiple languages into a normalized, structured JSON schema.

EVALUATION CRITERIA:
1. TRANSLATION & TOPIC: Translate core complaints into clear English. Classify topic into: `Billing`, `Logistics`, `Product_Defect`, or `Customer_Support`.
2. SENTIMENT & SEVERITY:
   - `sentiment_score`: Continuous float from -1.0 (severe dissatisfaction) to +1.0 (delighted).
   - `urgency_level`: Assign `P1` to financial/double-billing issues requiring immediate intervention, `P2` to delayed logistics, and `P3` to minor cosmetic flaws.
3. Output strictly inside <reviews_analysis> as a valid JSON array.
</system_instructions>