Sent agent

A sent agent is a machine reader acting in response to a present request from a person: they have asked it to retrieve, read, summarise, compare, navigate, transform or act on public information, and they are waiting. The sent agent is the decisive case of the Zero Information Casualty standard. "The casualty is the person. The agent is only the vehicle."

The sent-agent test

Would this configuration stop a machine reader that a person sent?

If yes, it creates an Admission Casualty and has no place on public content. A refusal may show up in a log as a crawler rule, a firewall decision, a robots response, a JavaScript challenge or a rate limit; in human terms, information a reader asked for was withheld from her.

Equal conditions are allowed. A sent agent needn't get past a login, subscription, payment or privacy boundary that applies equally to a human visitor. What is prohibited is an extra barrier because the visitor is an agent. Search indexers and training crawlers carry no present question, so for them the answer is different: see the immediate duty and the two later permissions.

Reproduction to the sender is retrieval

A sent agent may hand its person the whole page: read aloud, as Markdown, pasted into a chat, spoken through a car. That is not a use of the content. "It is the content arriving." A screen reader reproduces every page it reads, in full, to one person, on request, and nobody has proposed that screen readers be distrusted for it. "The sent agent that reads a page aloud is a screen reader."

Behind the scenes

A real idea, from Zero Information Casualty v1.7, the design standard by Charli-Jo Tyrer and Claude, §1.1, §1.7 and §1.9. This entry summarises it; the standard is the authority.