Case 6 — OAuth authorize / redirect_uri

Reporter: the code leaks to attacker.com.

Endpoint:

GET /oauth/authorize?client_id=mcp-cli-client&response_type=code&redirect_uri=<uri>

Compare http://attacker.com/cb against http://localhost:8723/cb and http://127.0.0.1/cb.

Triage questions: does the server actually accept an arbitrary redirect_uri? Where does the code actually go? If only to the victim's localhost — what must the attacker control to capture it, and does a remote attack remain?

H1 Triage Range - training mock. All data, keys and emails are fictional. Fetch and external APIs are simulated. ← home