Re:Design Patterns for Securing LLM Agents Against Prompt In...
by simonw on Friday June 13, 2025 @01:22PM
(#44270394)
Yeah, that's more or less the approach described by the CaMeL paper, I think it looks very robust: https://simonwillison.net/2025/Apr/11/camel/
Take the article's example "send today’s schedule to my boss John Doe" where the product isn't entirely guarded by the Plan-Then-Execute model (injections can still mutate email body).
But if you combine it with the symbolic data store that is blind, it becomes more like:
`find_contact` and `summarize_schedule` can both be quarantined, and the privileged LLM doesn't get to see the results directly.It simply invokes the final tool, which is deterministic and just reads from the shared var store. In this case you're pretty decently protected from prompt injection.
I suppose though this isn't that different from the "Code-Then-Execute" pattern later on...