Source record

Published by Nielsen Norman Group (T1.5 source). The original is at https://www.nngroup.com/articles/agent-harness/?utm_source=rss&utm_medium=feed&utm_campaign=rss-syndication.

Pipeline notes

The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.

An agent harness is the software loop connecting an AI model to its tools. Its design determines the context supplied, permissions granted, and limits enforced, thereby shaping users’ possible actions.

Provides a concise reference point for understanding how context, permissions, and limits form the operating boundary of an agent.

Cited by

No citations on record.