Some checks failed
CI / lint (push) Successful in 36s
Deploy / notebooks (push) Has been skipped
Deploy / zotero (push) Has been skipped
Deploy / docs (push) Has been skipped
Deploy / api (push) Has been skipped
Deploy / mc (push) Has been skipped
Deploy / report (push) Successful in 18s
CI / test (push) Failing after 12m53s
Before this change HTTPS_PROXY was a polite recommendation, not a boundary: proxy_net was a regular bridge with NAT, so anything inside the claude container could open a direct socket to the internet and silently bypass squid's allowlist. Now proxy_net is internal (no NAT, no outbound). cf-proxy joins a second bridge egress_net to dial the Cloudflare edge. claude stays on proxy_net only, so its only route off-host is through cf-proxy → CF Access → tunnel → squid → allowlisted CONNECTs. Verified: from a sibling container on proxy_net, curl https://1.1.1.1 fails immediately ("Could not connect to server"); the same container through cf-proxy gets HTTP 401 from api.anthropic.com (the smoke-test success code).