From 281e297db7ad040687ca10f45b5ffcc3c46fafb4 Mon Sep 17 00:00:00 2001
From: gbell27 <gabriele.bellavia.m@gmail.com>
Date: Wed, 25 Feb 2026 05:57:02 +0100
Subject: [PATCH] docs: add security warnings for WebFetch and FileSystem

---
 datapizza-ai-tools/filesystem/README.md | 22 ++++++++++++++++++++++
 datapizza-ai-tools/web_fetch/README.md  | 22 ++++++++++++++++++++++
 2 files changed, 44 insertions(+)

diff --git a/datapizza-ai-tools/filesystem/README.md b/datapizza-ai-tools/filesystem/README.md
index 0c8c7542..cf1ca7ea 100644
--- a/datapizza-ai-tools/filesystem/README.md
+++ b/datapizza-ai-tools/filesystem/README.md
@@ -196,3 +196,25 @@ Agent Response: Successfully deleted directory '/tmp/tmp_XXXXXX/initial_dir'.
 
 Cleaned up temporary directory: /tmp/tmp_XXXXXX (actual path will vary)
 ```
+
+> ⚠️ **Warning**: Paths are not automatically normalized. Malicious inputs like `../../../../etc/passwd` may bypass `paths_to_include`/`paths_to_exclude` filters, leading to **path traversal vulnerabilities**.
+
+**Best Practices**:
+- Normalize paths using `os.path.normpath` before any operation.
+- Restrict access to known-safe directories using `paths_to_include`.
+
+**Example: Safe Path Handling in Python**
+```python
+import os
+
+def safe_path(base_dir, user_path):
+    # Ensure the user-provided path stays within the base directory.
+    norm_path = os.path.normpath(os.path.join(base_dir, user_path))
+    if not norm_path.startswith(base_dir):
+        raise ValueError("Path traversal attempt detected")
+    return norm_path
+
+# Usage:
+base_dir = "/safe/directory"
+safe_user_path = safe_path(base_dir, user_input_path)
+```
\ No newline at end of file
diff --git a/datapizza-ai-tools/web_fetch/README.md b/datapizza-ai-tools/web_fetch/README.md
index 7657292c..7a364705 100644
--- a/datapizza-ai-tools/web_fetch/README.md
+++ b/datapizza-ai-tools/web_fetch/README.md
@@ -82,3 +82,25 @@ The output will vary depending on the live content of the URL. For `https://lore
 --- Running agent for: 'Summarize the main points of the article at https://loremipsum.io/' ---
 Agent Response: The article on **loremipsum.io** provides a comprehensive overview of "Lorem Ipsum," which is a placeholder text commonly used in the graphic, print, and publishing industries. Here are the main points:
 ```
+
+> ⚠️ **Warning**: This tool accepts unsanitized URLs. If exposed to untrusted input, it could be exploited for **Server-Side Request Forgery (SSRF)**, such as accessing internal services or cloud metadata endpoints (e.g., `http://169.254.169.254/`).
+
+**Best Practices**:
+- Always validate URLs before processing (e.g., allowlist trusted domains).
+- Avoid exposing this tool to uncontrolled input in production environments.
+- Use `paths_to_include` and `paths_to_exclude` to restrict access to sensitive resources.
+
+**Example: URL Validation in Python**
+```python
+from urllib.parse import urlparse
+
+def is_safe_url(url, allowed_domains):
+    # Check if a URL belongs to a list of allowed domains.
+    parsed = urlparse(url)
+    return parsed.netloc in allowed_domains
+
+# Usage:
+allowed_domains = ["api.trusted-service.com", "public-data.example.org"]
+if not is_safe_url(user_input_url, allowed_domains):
+    raise ValueError("URL not allowed")
+```
\ No newline at end of file