A dormant security flag that, when enabled, detects proxy interception of GCS artifact uploads by checking for a header that only real GCS responses include.
Feature-flagged (off by default)
- Export name
isFrameSignedPutHeaderCheckEnabledreveals the intent. - After a successful (2xx) HTTP PUT to Google Cloud Storage, the guard checks for the
x-goog-generationresponse header — present on every authentic GCS response, absent on proxied ones. - If the header is missing on a 2xx, the upload is rejected with:
"upload intercepted: 2xx without x-goog-generation — a proxy answered in place of GCS"(status 403,intercepted: true). - Scoped to GCS PUT responses only. Anthropic API traffic (
api.anthropic.com), telemetry, OAuth, and MCP are unaffected — this does not block proxying of model requests or conversation content. - When enabled, would break transparent proxies that cannot forward GCS headers intact: Burp Suite, mitmproxy, corporate SSL inspection, LLM gateways (LiteLLM, Portkey, etc.).
Flag check mrl() in vKp() GCS upload handler; export isFrameSignedPutHeaderCheckEnabled (search for "x-goog-generation")
Strings lifted out of the shipped bundle, so the claim above can be checked against them.
tengu_cobalt_plinth_putguard Off by default, switched on for this accountThe shipped code defaults this off, and the flag server returned on for the one account this site reads on this version. That is the reading that makes the entry above worth a second look, and it still says nothing about your account.
This account: on · anonymous baseline: on · compiled default in v2.1.193: off
These values were read against a different version of Claude Code, so treat them as the nearest reading available instead of one taken on this release.
Read once, for one account on one subscription tier, against v2.1.193. It isn't a statement about your account. What a flag value here can and cannot tell you
A reading is one sample. Claude Code evaluates its flags remotely, so no client sees the targeting rule behind a value and this says nothing about your account.
A reading is one sample. Claude Code evaluates its flags remotely, so no client sees the targeting rule behind a value and this says nothing about your account.
New in this build: tengu_cobalt_plinth_putguard