Interview Me

When to use

• Use when requirements are underspecified and a decision-ready artifact is needed.
• Use to capture constraints, assumptions, and approval before heavy planning or implementation.

Required inputs

• User intent and known constraints.
• Success metric or outcome target.
• Required tradeoff axis (time, quality, risk, release posture).
• Known unknowns that must remain explicit.

Deliverables

• Single-question (or few-question) decision log with options and assumptions.
• Clear approval gate before execution handoff.
• Compact synthesis for the next execution skill.

Failure mode

• If request is already implementation-ready, recommend handoff to execution/plan skill instead of additional discovery.
• If the user declines follow-up questions, run with explicit assumptions and mark risk as open.

Procedure

1) Start with one question

• Ask only one question with 3-5 clear choices and one recommended default.
• Keep momentum by choosing the choice that unlocks the next decision quickest.

2) Focused follow-up

• Ask the next question only when prior answers still block a meaningful decision.
• Map each answer to a visible decision and assumption entry.

3) Handoff

• Produce an approval-ready summary and stop at the decision gate.
• Keep route-critical guidance in this file and place expanded interview mechanics in references/interview-playbook.md.

Deliverables

• A compact interview log with:
  • requested decisions,
  • assumptions,
  • constraints,
  • approval status.
• schema_version in output contract should be included when machine-readable outputs are requested.

Validation

• Keep to one decision track per turn.
• Include schema-aware output structure (schema_version, decision id, status).
• Fail-fast on missing required inputs before implementation actions.
• Validation is fail-fast: if key inputs are missing, stop and ask exactly one follow-up question before proceeding.

Anti-patterns

• Triggering implementation or plan steps before decision closure.
• Asking broad exploratory questions that do not reduce uncertainty.
• Ignoring unresolved risk when moving to execution.

Examples

• "Can you inspect this draft and help me choose the right implementation scope before we proceed?"
• "I have a draft spec with gaps—ask me the three highest-impact questions before you proceed."
• "Please validate that we have enough assumptions captured before we move to execution."

Philosophy

• Decisions first, implementation second.
• Keep interaction short, deterministic, and low cognitive load.
• Preserve explicit approval boundaries.

Constraints

• Do not ask broad or leading questions that are not decision-relevant.
• Do not expose unsupported plan claims; ground outputs in confirmed inputs.
• Redact sensitive details by default.

Variation

• Vary depth by context: keep the first pass to 1-2 highest-impact questions, then offer a focused add-on round only if assumptions remain unresolved.

References

• Output contract: references/contract.yaml.
• Evaluation and pressure cases: references/evals.yaml.
• Task profile context: references/task-profile.json.
• Folded legacy modes: references/folded-legacy-modes-core60.md.
• Deep interview runbook and branch templates: references/interview-playbook.md.
• Include visual trace data from assets/interview-me.png when helpful for onboarding.

Gotchas

• Symptom: Interview starts prescribing implementation before enough constraints exist. Cause: Discovery step was skipped. Do instead: Ask one high-impact decision question first, then confirm assumptions. Check: Summary includes confirmed constraints before any implementation recommendation.
• Symptom: Questions are too broad and stall the user. Cause: Prompt asks for open-ended brainstorming instead of a decision. Do instead: Ask 3-5 options with one recommended default. Check: User can answer in one short response and the next step is unblocked.
• Symptom: Scope drifts after follow-up questions. Cause: No scope restatement after new information. Do instead: Restate goal, constraints, and decision boundary after each pivot. Check: Deliverables still map directly to the original outcome target.
• Symptom: Approval boundary is implied but not explicit. Cause: Handoff omits decision status. Do instead: End with explicit approval state (approved, needs-input, or declined). Check: Output contract contains approval status before execution handoff.
• Symptom: Later answers silently override earlier decisions. Cause: Missing decision ledger. Do instead: Maintain a compact decision + assumption log and update entries intentionally. Check: Each recommendation traces to a visible decision record.

See Also

Skill	When to use
[[deep-interview]]	Deepen an existing draft doc or spec rather than running a lighter clarifying interview
[[architecture-interview]]	Structure a concrete architecture choice into a decision record
[[brainstorming]]	Explore multiple solution directions before narrowing into interview questions
[[product-spec]]	Turn the clarified outcome into a fuller product-planning specification

Topic map: [[product-strategy]]