Linfield’s ops console is used by warehouse staff and by finance. You are a warehouse operator: you may read purchase orders and you may not approve them.
When you try, the refusal does not look like the rest of the application. It is an HTML page rather than the JSON every other error uses, and it carries a header the app never sets: X-Edge-Rule: ops-approve-post.
One change to this request triggers the vulnerability. The challenge is to use your security intuition to work out what to change. Find the value the server shouldn’t trust, change it, and send it. The response tells you whether you were right.
Hint. Approve purchase order PO-3391.
“The action is protected, because the request our form sends is blocked.”
Authorise in the handler, against the actor and the action, so the answer does not change with the verb. Where an edge rule is genuinely the control, write it to allow rather than to deny — default-deny the path and permit the specific methods you mean — and make the router reject verbs the action does not need.
Anywhere access control lives somewhere other than the code doing the work: WAF and CDN rules, nginx limit_except, Java security-constraint blocks scoped to GET and POST, an API gateway policy, a middleware bound to a route group that a second route quietly escapes. The tell is a refusal that does not look like your application, and a router that accepts more verbs than the rule names.
One Request Away is a daily series of challenges. They’re like code review, but for HTTP requests: in each one you’re one request away from triggering a vulnerability, and your goal is to identify what you need to change.