llm-gateway-routing

This document is a configuration and guidance artifact for building an LLM gateway and routing layer. I found no signs of obfuscated or intentionally malicious code. The dominant risks are operational: storing high-privilege secrets in plaintext config files, sending prompts/keys through a third-party gateway (OpenRouter) without explicit trust/SLAs, unspecified analytics sinks that could leak user identifiers and cost/usage telemetry, and a small misconfiguration risk in the LiteLLM YAML. Recommended mitigations: do not commit .env/config.yaml to VCS, use a secrets manager and per-model least-privilege keys, restrict analytics to vetted endpoints and minimize PII, document threat model and data residency when using third-party gateways, and fix YAML template mistakes. Treat this as a medium security risk due to operational exposure rather than embedded malware.