deep-interview

To convert vague requirements into clear specifications before autonomous execution, run a Socratic one-question-at-a-time interview with explicit ambiguity scoring. This exposes hidden assumptions and unstable scope, refusing to proceed until clarity thresholds are met.

Ambiguity scoring in requirements gathering mathematically evaluates goals, constraints, and success criteria to prevent building the wrong thing. It gates autonomous execution by refusing to proceed until computed ambiguity from weighted clarity dimensions drops to an acceptable threshold.

To gather repository facts for brownfield integration before writing a feature specification, perform brownfield-safe context checks via repo exploration. This confirms how the feature should fit the existing codebase, reducing wrong integration outcomes before downstream pipeline stages.

Yes, you can use Socratic questioning to gate autonomous execution for greenfield tasks with unclear success metrics. It applies to both greenfield and brownfield tasks, asking one question at a time to expose hidden assumptions and compute a readiness-gating ambiguity score.

The best way to define non-goals and measurable acceptance criteria from a vague idea is to run a Socratic deep interview that persists state and outputs a spec. It turns uncertainty into explicit, testable requirements with clear non-goals for downstream pipeline stages.

You should not use an autonomous agent for requirements gathering when your task lacks hidden assumptions or unstable scope, or when your goal, constraints, and success metrics are already explicit. In these cases, Socratic questioning and ambiguity scoring add unnecessary overhead.