What does Shield do?

Gnesi

A check before your agent reads anything

Shield inspects every incoming message written by someone else before an agent acts on it. It looks for prompt injection, hidden text, and requests that push an agent to ignore its instructions, forward mail, make payments or change account data. Shield is always on and cannot be turned off.

Verdicts
  • Clear: nothing manipulative was found. This is advice, not a guarantee.
  • Review: the message might be trying to manipulate an agent.
  • Unavailable: Shield could not score it, for example an empty or very large message.
Settings

Under Settings → Agents → Shield:

  • Keep a review away from my agent until I release it keeps flagged messages in Needs you. You can still read them yourself.
  • Also hold a polite request to forward, pay, or change data flags those requests too.

Only a signed-in person can release a held message. Agent credentials cannot.

Read more in the Gnesi docs.

Not finding what you're looking for ?

Check the Gnesi docs, or reach out and we will get back to you quickly.

Contact us