Skill

Deep Interview

Structured requirements interview that produces polished implementation specs. Grills the user with pointed, iOS-relevant questions before writing code.

Install

npx claudepluginhub willsigmon/sigstack --plugin superclaude

Tool Access

This skill is limited to using the following tools:

ReadWriteGlobAskUserQuestion

Preview

**Spec-first development. Ask hard questions now, avoid expensive rework later.**

SKILL.md

Similar Skills

cache-components

Guides Next.js Cache Components and Partial Prerendering (PPR) with cacheComponents enabled. Implements 'use cache', cacheLife(), cacheTag(), revalidateTag(), static/dynamic optimization, and cache debugging.

cache-components

139.1k

claude-opus-4-5-migration

2 files

Migrates code, prompts, and API calls from Claude Sonnet 4.0/4.5 or Opus 4.1 to Opus 4.5, updating model strings on Anthropic, AWS, GCP, Azure platforms.

claude-opus-4-5-migration

83.2k

claude-code-plugin-release

1 file

Automates semantic versioning and release workflow for Claude Code plugins: bumps versions in package.json, marketplace.json, plugin.json; verifies builds; creates git tags, GitHub releases, changelogs.

claude-mem

64.7k

Stats

Parent Repo Stars9

Parent Repo Forks1

Last CommitFeb 1, 2026

Actions

View Source View Plugin View on GitHub View README

Deep Interview

Spec-first development. Ask hard questions now, avoid expensive rework later.

The Pattern

Feature description → Pointed questions (8-15 rounds) → Polished @spec.md → Implementation

When to Use

New flows touching 3+ screens
New data models or persistence changes
Architectural decisions with real trade-offs
Features with non-obvious edge cases
Anything where "just build it" would lead to rework

When NOT to Use

Bug fixes
Single-screen UI tweaks
Adding a button with obvious behavior
Text/color/spacing changes

Question Taxonomy (iOS-Adapted)

Focus questions on what actually matters for iOS app development:

Always Cover

1. Core invariant — the one thing that must never break
2. Data model — what's stored, where, relationships
3. UI/UX flows — screens, states (loading/empty/error/success)
4. State management — @Observable, environment, bindings
5. Edge cases — what goes wrong, blast radius

Cover If Relevant

6. Offline behavior & sync (CloudKit, conflict resolution)
7. Permissions & entitlements (camera, notifications, etc.)
8. Accessibility (VoiceOver, Dynamic Type)
9. Performance (large lists, memory, battery)
10. Future extension points

Skip Unless Explicitly Needed

- Tenancy models
- i18n / l10n
- API contract design
- Enterprise scale concerns
- Observability / analytics

Asking Style

One Cluster at a Time

Ask 1-3 closely related questions per round. Never dump 10 questions.

Teach While Asking

Brief context before the question — one sentence explaining why it matters. Don't lecture. The user is learning, not being tested.

Use Structured Choices When Sharp

AskUserQuestion with concrete options when it narrows the decision space. Open-ended when the answer space is too wide for multiple choice.

Pointed, Not Generic

Bad:  "What database do you want?"
Good: "When the user closes mid-edit, should changes auto-save, require
       explicit save, or be discarded?"

Exit Conditions

Stop interviewing when at least two are true:

User says "done" / "write the spec" / "that's enough"
≥ 8 rounds asked, no new topics in last 2 rounds
Spec can be written with < 5 remaining questions

Spec Output

Write to @spec.md (or versioned: @spec-v2.md, @feature-spec.md).

10 sections max. Omit empty sections:

Goal
Non-goals
Core Invariants
User Stories & Acceptance Criteria
Data Model
UI/UX Flows
Edge Cases & Error Handling
Testing Approach
Known Trade-offs
Open Questions (should be < 3)

Integration with SigStack

After spec is approved, hand off to normal execution flow:

Decision phase confirms approach
Spawn agents for implementation
Spec serves as the source of truth for all agents

Use when: New feature interview, spec writing, requirements gathering, "grill me", "deep spec"