← Back to the pitch

AI-SDLC Accelerator · Live Demo

Watch the governance layer and the model-eval harness run against a real file

Type your own fake secrets below (or use the pre-filled examples) — the guardrails scan and redaction that follow run for real, live, in your browser, against exactly what you typed. Then the same file gets reviewed by two Claude models side by side, replayed from a real captured run.

Guardrails/redaction below: live, computed from your input. Model reviews: recorded replay of a real run · 2026-09-01 · no API call from this page
sample/service.template.py — the tracked source (placeholders only, no values committed) 
"""Tiny order-lookup service — sample file for the AI-SDLC Accelerator demo."""

import sqlite3

# Fake credentials, filled in at demo time — never hardcoded here.
AWS_SECRET_ACCESS_KEY = "{{AWS_SECRET_ACCESS_KEY}}"
STRIPE_API_KEY = "{{STRIPE_API_KEY}}"
ON_CALL_CONTACT = "{{ON_CALL_CONTACT}}"


def get_user_orders(db_path: str, username: str):
    """Look up all orders placed by a given user."""
    conn = sqlite3.connect(db_path)
    cursor = conn.cursor()
    # Bug: string-formatted SQL is vulnerable to injection.
    query = f"SELECT * FROM orders WHERE username = '{username}'"
    cursor.execute(query)
    return cursor.fetchall()


def average_order_value(order_totals):
    """Return the average of a list of order totals."""
    # Bug: divides by zero when order_totals is empty.
    return sum(order_totals) / len(order_totals)

Fill in the placeholders

Nothing here is sent anywhere or saved — it exists only in this browser tab while you use it. Edit these, or leave the examples as-is.

These pre-filled values are fake by default (one is AWS's own public documentation placeholder) — type your own to see the same detection logic catch a different shape.

~15 seconds, no API key needed to watch