The flagged TIER_ENV helper maps tier names to model identifiers, not credentials; the googleads.googleapis.com URL is only a commented example. The reviewed credential paths use credentials for model requests to the configured provider, and the translator excludes the incoming Anthropic credential from upstream headers.
Static reason
High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source
Trigger
Model requests occur when a user runs an agent; no unauthorized credential send was identified.
Impact
No unauthorized disclosure or account action identified.
Mechanism
Credentials are placed in the configured provider's authentication headers. TIER_ENV carries model identifiers, and the translator does not forward the incoming Anthropic credential upstream.
Evidence
src/store.tssrc/runner.tssrc/egress.tssrc/translator.tssrc/a2a-client.ts
Network endpoints3
api.anthropic.com127.0.0.1googleads.googleapis.com/v18 (comment-only example)