Deep Interview

Turn incomplete intent into a user-approved Interview Brief. Interview first; do not implement the resulting work in the same invocation unless the user separately authorizes that next step.

Quick Start

1. Ground on the conversation and in-scope files or prior artifacts. Do not make the user repeat inspectable context.
2. Choose the opening mode:
   • Sparse context: begin Socratic discovery with the highest-impact unresolved decision.
   • Rich context: begin with an inverted interview: synthesize confirmed facts, inferred assumptions, unresolved ambiguities, and human-only decisions, then ask which numbered item is wrong or incomplete.
   • Brownfield work: inspect relevant code and docs before asking; cite the evidence that makes a user decision necessary.
3. Maintain an ambiguity ledger for intent, actors, outcome, current facts, scope, non-goals, constraints, authority, verification, risks, and decisions.
4. Route every uncertainty before asking: inspect discoverable facts, confirm conflicting or interpretive facts, and ask the user only for judgment.
5. Ask exactly one focused question per turn and wait. Target the unresolved answer most likely to change the outcome or downstream execution.
6. Pressure-test at least one material answer with an example, assumption, failure scenario, counterexample, or tradeoff.
7. Run the closure gate, restate the intended outcome in one sentence, and ask the user to confirm or correct it unless they already accepted that same outcome and the material decisions have not changed.
8. After confirmation, produce the Interview Brief using assets/interview-brief.template.md. Return it in chat by default; write a file only when the user asks for one.

Read references/interview-method.md for detailed lenses and examples, references/harness-source-notes.md when maintaining the skill, and references/evaluation-cases.md when changing behavior.