Skip to main content
Comet automates safe transitions, but product and risk decisions still belong to you. A decision point is a hard pause where the agent must get an explicit answer before continuing.

Xiaoyu waits at decision signposts along the five-phase Comet path

Decide whether user input is actually required

Since 0.4.0-beta.5, Comet first classifies every situation before pausing:
  • User decision — two or more valid options change scope, behavior, accepted risk, or an irreversible outcome; you must choose.
  • Automatic handling — exactly one safe next action remains (repair an objective failure, reconcile verifiable state, retry an idempotent check, follow persisted configuration); execute and report it without manufacturing confirmation.
  • Stop condition — a missing dependency, corrupt state, path escape, or unavailable external command leaves no valid next action; report the blocker and recovery condition.
  • Manual handoffNEXT: manual returns control; it is not a new user decision point. Print HINT, end the current invocation, and do not ask whether to continue.
Only the first category uses the blocking rules below.

Rules

  • The agent cannot replace your answer with a default.
  • A recommendation is not a decision.
  • A past preference is not a decision.
  • Text output alone is not enough when the platform has a structured question UI.
  • Merge adjacent choices that can be answered together, and do not re-ask persisted choices that remain valid.

Common decision points

Why this matters

Comet is designed to be recoverable and auditable. Silent defaults at risk points make later recovery misleading.

Next steps

Last modified on July 24, 2026