Creating Your First Prompt

Build a code review prompt from scratch.

Design

A code review prompt needs:

Code snippet input
Programming language
Review style option

Create the YAML

Create code_review.yaml:

name: code_review
description: AI-powered code review
template: |
  Review the following {{ language }} code:

  ```{{ language }}
  {{ code }}

{% if style == “strict” %} Focus on bugs, performance, and style violations. Be critical. {% elif style == “security” %} Focus on security vulnerabilities and attack vectors. {% else %} Provide helpful feedback on readability and best practices. {% endif %}

Format your review as:

Summary: Overall assessment
Issues: Problems found
Suggestions: Improvements

input_schema: code: str language: str style: “str | None”

## Use in Python

```python
from promptkit import load_prompt, run_prompt, OpenAIEngine

prompt = load_prompt("code_review.yaml")

code = """
def calculate(items):
    total = 0
    for item in items:
        total += item.price
    return total
"""

engine = OpenAIEngine(api_key="sk-...")
review = run_prompt(
    prompt,
    {"code": code, "language": "python", "style": "helpful"},
    engine
)
print(review)

Use the CLI

# Render to preview
promptkit render code_review.yaml \
  --vars '{"code": "def hello(): pass", "language": "python"}'

# Validate
promptkit lint code_review.yaml

# Get info
promptkit info code_review.yaml

Best Practices

Clear Schema

input_schema:
  user_id: str                  # Required
  preferences: "list | None"    # Optional

Default Values

Use Jinja2 filters:

template: |
  Hello {{ name | default("there") }}!
  Priority: {{ priority | default("normal") }}

Conditional Logic

template: |
  {% if detailed %}
  Provide comprehensive analysis.
  {% else %}
  Provide brief summary.
  {% endif %}