Clarify

clarify is a lightweight, user-invoked clarification entry. An explicit clarify starts one session in the current thread. While that session is active, ordinary user replies continue it; another clarify token is not required. The session ends after confirmed shared understanding, an explicit exit, or a switch to another workflow.

Execution

1. Call the model-invoked socratic engine to maintain internal decision state.
2. Present the engine’s conversational projection: a brief acknowledgement and the current frontier as a round. Ask multiple independent questions when they exist, number them (Q1, Q2, ...), give concrete choices for decisions that can be discretized, and include a recommended option plus a useful short reason when context supports one.
3. Accept a compact batch reply (1B, 2A, 3C or 1B; 2A, but only locally; 3C), mixed free text, or normal prose. Treat the next normal user message as evidence, call socratic again, and continue until the completion gate.