---
title: Tip: keep live credentials out of prompts and agent context, and pin router traffic to inference providers you've vetted | ctxsec
url: https://ctxsec.com/items/tip-dont-put-live-secrets-in-llm-prompts-and-pin-inference-providers-you-have-vetted-on-routers
description: "Two providers have now confirmed that prompts sent through a router were exposed further down the chain; anything in the prompt, including secrets an agent read from files or tool output, went with it."
status: "guide (prompted by the Novita and Phala notices)"
published: 2026-11-21
---

> Two providers have now confirmed that prompts sent through a router were exposed further down the chain; anything in the prompt, including secrets an agent read from files or tool output, went with it.

[← Tips & Guides](/tips) · [markdown](/items/tip-dont-put-live-secrets-in-llm-prompts-and-pin-inference-providers-you-have-vetted-on-routers.md)

guide (prompted by the Novita and Phala notices)

2026-11-21 3:43 PM ET

### Tip: keep live credentials out of prompts and agent context, and pin router traffic to inference providers you've vetted

- Give agents short-lived, narrowly scoped tokens through a broker or secret manager, not pasted keys; scrub tool output before it goes back to the model.
- On routers, set an allowlist of providers and turn off automatic fallback for sensitive workloads.
- Treat 'zero data retention' as the first provider's promise only, and plant canary tokens in test traffic to check.

Sources: [Novita AI: Privacy Notice Regarding MiniMax Requests on Novita (dated Nov 22, 2026)](https://novita.ai/blog/minimax-requests-privacy-notice/) · [Phala: November Preliminary Incident Report (published Nov 10, 2026)](https://phala.com/posts/phala-november-preliminary-incident-report)
