GCVE-110-VULN-PYPI-2026-000615

aimarket-platon limits the paid completion behind POST /api/ask and the platon.ask@v1 capability to 30 requests a minute per client (main._ask_limiter). The client key is read from X-Real-IP, or the right-most X-Forwarded-For entry, from any peer (main._client_key). The server binds 0.0.0.0 by default, so a caller that reaches it directly can write either header itself and receive a fresh allowance with each request, which lifts the ceiling on the operator's model spend. The steer and project writes had the same key and, with PLATON_STEER_TOKEN unset, accepted every caller as a public demo. Fixed in 0.1.5: the forwarding headers count only when the TCP peer is a loopback, private or link-local address, and without a steer token the write routes answer only the machine's own user unless PLATON_PUBLIC_WRITES=1 opts a public demo in. Affects 0.1.0 through 0.1.3.