Ask HN: How are you controlling AI agents that take real actions?
thesvp Tuesday, February 24, 2026We're building AI agents that take real actions — refunds, database writes, API calls.
Prompt instructions like "never do X" don't hold up. LLMs ignore them when context is long or users push hard.
Curious how others are handling this: - Hard-coded checks before every action? - Some middleware layer? - Just hoping for the best?
We built a control layer for this — different methods for structured data, unstructured outputs, and guardrails (https://limits.dev). Genuinely want to learn how others approach it.
2
8