← sources

Need ways to guardrail my agents

source: reddit (reddit:cybersecurity)published 2026-11-09 04:55ingested 2026-11-10 02:38

retained text

I’m looking to give my ai more permissions to play around with reading online documents and doing some investing through Robinhood’s mcp connection. What do people use to protect against prompt injection or potential mishaps while it is running?

TLDR: Scared of AI going off prompt, looking for solutions

document fetched 2026-11-10 00:23 · canonical https://reddit.com/r/cybersecurity/comments/1xrit5q/need_ways_to_guardrail_my_agents