claude-code-basic-ai-detector/app.py at main · Zshader/claude-code-basic-ai-detector · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
"""Starter AI Defense Lab: a tiny web app to try out prompt injection defenses.

Run it with:
    python app.py

Then open http://127.0.0.1:5000 and submit some text to see whether the
detector flags it as a possible prompt injection attempt.
"""

from flask import Flask, render_template_string, request

from src.detector import categorize, is_suspicious, sanitize, scan

app = Flask(__name__)

PAGE = """
<!doctype html>
<title>AI Defense Lab</title>
<h1>AI Defense Lab</h1>
<p>Paste some text below to check it for common prompt injection patterns.</p>
<form method="post">
  <textarea name="text" rows="6" cols="60" placeholder="Type or paste text here...">{{ text }}</textarea><br>
  <button type="submit">Check</button>
</form>
{% if checked %}
  <h2>Result</h2>
  <p><strong>Suspicious:</strong> {{ suspicious }}</p>
  <p><strong>Matched phrases:</strong> {{ matches }}</p>
  <p><strong>Attack categories:</strong> {{ categories }}</p>
  <p><strong>Sanitized text:</strong> {{ sanitized }}</p>
{% endif %}
"""


@app.route("/", methods=["GET", "POST"])
def index():
    text = ""
    checked = False
    suspicious = False
    matches = []
    categories = []
    sanitized = ""

    if request.method == "POST":
        text = request.form.get("text", "")
        checked = True
        matches = scan(text)
        categories = categorize(text)
        suspicious = is_suspicious(text)
        sanitized = sanitize(text)

    return render_template_string(
        PAGE,
        text=text,
        checked=checked,
        suspicious=suspicious,
        matches=matches,
        categories=categories,
        sanitized=sanitized,
    )


if __name__ == "__main__":
    app.run(debug=True)