What happened
career-ops's local dashboard fails to validate the Origin header or restrict its command-execution and file-write API endpoints to loopback-only access, allowing a malicious web page to trigger command execution on the host of any user running the dashboard, a classic local-service CSRF-to-RCE pattern.
Why it matters
This is representative of a recurring class of vulnerability in 'local AI assistant' tools that expose unauthenticated localhost APIs — a pattern flagged elsewhere in this same reporting window (Windows ML CLI, similar AI dev tools) — where a user simply browsing the web while the tool runs can have their machine compromised.
Attack vector
The local web dashboard exposes command-spawning and file-writing /api routes without validating request origin or restricting to loopback-only clients; a malicious webpage visited by the user while the dashboard is running can issue cross-origin requests that spawn commands or write files on the user's machine.
Affected systems
career-ops, versions prior to 0.8.0
Mitigation
Upgrade to career-ops 0.8.0 or later, which adds origin validation/loopback restriction to the local dashboard API.