Generating Test Cases with LLMs

You are a senior QA engineer. Design test cases for the requirement below.

Requirement:
A discount code field at checkout. Codes are 6–12 characters, letters and
digits only, case-insensitive. A code can be used once per customer, only
before its expiry date, and only on orders of at least $20. Only one code
per order. The discount can never make the total negative.

Instructions:
- First list any ambiguities or missing rules as questions. Do not invent answers.
- Apply equivalence partitioning, boundary value analysis and a decision table.
- Include positive, negative, boundary, security and usability cases.
- Mark each case with priority P1 (must run every release) to P3.

Output a Markdown table with columns:
ID | Title | Technique | Preconditions | Steps | Test data | Expected result | Priority

pip install anthropic
export ANTHROPIC_API_KEY="your-key"     # Windows PowerShell: $env:ANTHROPIC_API_KEY="your-key"

# generate_tests.py — run with: python generate_tests.py
import csv
import json

import anthropic

REQUIREMENT = """Discount codes are 6-12 letters/digits, case-insensitive, single use per
customer, valid until their expiry date, for orders of at least $20, one code per order,
and the total can never go below $0."""

PROMPT = f"""You are a senior QA engineer. Design test cases for this requirement:

{REQUIREMENT}

Apply equivalence partitioning, boundary value analysis and decision tables. Include
negative and security cases. Do not invent rules that are not in the requirement.

Respond with JSON only, no prose, in this shape:
{{"questions": ["..."],
  "test_cases": [{{"id": "TC-01", "title": "...", "technique": "...",
                  "steps": "...", "test_data": "...", "expected": "...", "priority": "P1"}}]}}"""

client = anthropic.Anthropic()     # reads ANTHROPIC_API_KEY from the environment
message = client.messages.create(
    model="claude-opus-5-5",
    max_tokens=4000,
    messages=[{"role": "user", "content": PROMPT}],
)

FENCE = "`" * 3                                  # a Markdown code fence
text = message.content[0].text.strip()
if text.startswith(FENCE):                       # tolerate a fenced JSON block
    text = text.strip("`").removeprefix("json").strip()
result = json.loads(text)

print("Open questions for the product owner:")
for q in result["questions"]:
    print(" -", q)

with open("generated_test_cases.csv", "w", newline="", encoding="utf-8") as f:
    writer = csv.DictWriter(f, fieldnames=["id", "title", "technique", "steps", "test_data", "expected", "priority"])
    writer.writeheader()
    writer.writerows(result["test_cases"])

print(f"\nWrote {len(result['test_cases'])} draft test cases to generated_test_cases.csv — review before use.")

Generate 15 values for a "full name" field to test a registration form.
Include: very short and very long names, apostrophes and hyphens,
Urdu and Arabic script, names with diacritics, a single word, leading
and trailing spaces, an emoji, and an HTML/script injection attempt.
Return one value per line with a short note on what each tests.

# pip install faker
from faker import Faker

fake = Faker(["en_US", "ar_AA"])
Faker.seed(42)                      # same data every run → reproducible tests
for _ in range(3):
    print(fake.name(), "|", fake.email(), "|", fake.phone_number())