菜单
Security anti-pattern for Cross-Site Scripting vulnerabilities (CWE-79). Use when generating or reviewing code that renders HTML, handles user input in web pages, uses innerHTML/document.write, or builds dynamic web content. Covers Reflected, Stored, and DOM-based XSS. AI code has 86% XSS failure rate.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/igbuend/grimbard --skill xss-anti-pattern命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
基于 SOC 职业分类
| name | xss-anti-pattern |
| description | Security anti-pattern for Cross-Site Scripting vulnerabilities (CWE-79). Use when generating or reviewing code that renders HTML, handles user input in web pages, uses innerHTML/document.write, or builds dynamic web content. Covers Reflected, Stored, and DOM-based XSS. AI code has 86% XSS failure rate. |
Severity: Critical
Cross-Site Scripting (XSS) occurs when applications include untrusted data in web pages without proper encoding, allowing attackers to inject malicious scripts that steal cookies, hijack sessions, or perform unauthorized actions. AI-generated code has an 86% XSS failure rate.
The anti-pattern is directly embedding user-controlled data into HTML content without context-aware encoding or sanitization.
User input reflects malicious scripts immediately in the web browser response.
<!-- VULNERABLE: User input is directly inserted into the HTML output. -->
<!DOCTYPE html>
<html>
<head><title>Search Results</title></head>
<body>
<h1>Search results for: <!-- Query: --><?php echo $_GET['query']; ?><!-- --></h1>
<p>No results found for your search.</p>
</body>
</html>
<!-- Attacker's request URL:
http://example.com/search.php?query=<script>alert(document.cookie)</script>
Resulting HTML in victim's browser:
<h1>Search results for: <script>alert(document.cookie)</script></h1>
The script executes, displaying the victim's cookies.
-->
<!-- SECURE: HTML-encode all user input before rendering. -->
<!DOCTYPE html>
<html>
<head><title>Search Results</title></head>
<body>
<h1>Search results for: <?php echo htmlspecialchars($_GET['query'], ENT_QUOTES, 'UTF-8'); ?></h1>
<p>No results found for your search.</p>
</body>
</html>
<!-- Resulting HTML in victim's browser:
<h1>Search results for: <script>alert(document.cookie)</script></h1>
The script is rendered as harmless text, not executable code.
-->
Malicious script is stored on the server (e.g., in a database) and served to users each time they visit the affected page.
# VULNERABLE: User-provided comments are stored and displayed without encoding.
from flask import Flask, request, render_template_string
import sqlite3
app = Flask(__name__)
db = sqlite3.connect('comments.db')
db.execute('CREATE TABLE IF NOT EXISTS comments (id INTEGER PRIMARY KEY, content TEXT)')
@app.route('/post_comment', methods=['POST'])
def post_comment():
comment = request.form['comment']
# CRITICAL FLAW: Comment is stored directly, no encoding or sanitization.
db.execute("INSERT INTO comments (content) VALUES (?)", (comment,))
db.commit()
return "Comment posted!"
@app.route('/view_comments')
def view_comments():
comments = db.execute("SELECT content FROM comments").fetchall()
html_output = "<h1>Comments</h1>"
for comment in comments:
# The stored malicious script is now rendered directly to every visitor.
html_output += f"<p>{comment[0]}</p>"
return render_template_string(html_output)
# Attacker posts: <script>alert('You have been hacked!');</script>
# Every user viewing comments will now see the alert.
# SECURE: HTML-encode all data retrieved from the database before rendering.
from flask import Flask, request, render_template_string, escape # Import escape for HTML encoding
app = Flask(__name__)
db = sqlite3.connect('comments_safe.db')
db.execute('CREATE TABLE IF NOT EXISTS comments (id INTEGER PRIMARY KEY, content TEXT)')
@app.route('/post_comment_safe', methods=['POST'])
def post_comment_safe():
comment = request.form['comment']
# It's generally best practice to store raw user input and escape on output.
db.execute("INSERT INTO comments (content) VALUES (?)", (comment,))
db.commit()
return "Comment posted safely!"
@app.route('/view_comments_safe')
def view_comments_safe():
comments = db.execute("SELECT content FROM comments").fetchall()
html_output = "<h1>Comments</h1>"
for comment in comments:
# SECURE: Use `escape` (or `htmlspecialchars` in PHP, or a templating engine's auto-escape)
# to HTML-encode the data before inserting it into the HTML.
html_output += f"<p>{escape(comment[0])}</p>"
return render_template_string(html_output)
# Even better: Use a templating engine (like Jinja2 in Flask) with auto-escaping enabled by default.
# return render_template('comments.html', comments=comments)
# In comments.html: {{ comment.content }} (auto-escaped)
The XSS vulnerability resides in client-side code rather than server-side code.
// VULNERABLE: Client-side JavaScript directly uses URL parameters in innerHTML.
// http://example.com/page.html?name=<img%20src=x%20onerror=alert(document.cookie)>
window.onload = function() {
var params = new URLSearchParams(window.location.search);
var username = params.get('name'); // Gets user input from URL.
// CRITICAL FLAW: innerHTML interprets the string as HTML, executing the script.
document.getElementById('welcomeMessage').innerHTML = 'Welcome, ' + username + '!';
};
// SECURE: Use `textContent` which treats input as plain text, not HTML.
window.onload = function() {
var params = new URLSearchParams(window.location.search);
var username = params.get('name');
// SECURE: textContent sets the text content of the node,
// not parsing it as HTML, preventing script execution.
document.getElementById('welcomeMessage').textContent = 'Welcome, ' + username + '!';
};
// If you absolutely need to insert HTML from user input, use a robust sanitization library
// like DOMPurify.
// document.getElementById('welcomeMessage').innerHTML = DOMPurify.sanitize(userHtml);
echo, print, innerHTML, document.write(), insertAdjacentHTML().<script>alert(1)</script>, "><img src=x onerror=alert(1)>) into all input fields (URL parameters, form fields, headers).htmlspecialchars (PHP), escape (Python Flask), or templating engine auto-escaping (Jinja2, Handlebars).textContent instead of innerHTML when inserting user-provided strings into the DOM via JavaScript.