How does Deep Interview differ from a standard AI chat?

While standard chat often makes assumptions to keep moving, Deep Interview identifies what it doesn't know and asks targeted, one-at-a-time questions to expose hidden assumptions.

Can I use Deep Interview on existing projects?

Yes. It features a 'brownfield' mode that uses a haiku-powered agent to explore your current codebase and ensure new requirements fit existing patterns.

What is mathematical ambiguity gating?

It is a scoring system that evaluates project clarity across dimensions like goals and constraints. The skill requires this score to reach a threshold (usually 80% clarity) before it proceeds to build anything.

Is it possible to skip the interview if I'm in a hurry?

Yes. While designed for thoroughness, the skill allows for early exits after round 3 if you decide the current level of clarity is sufficient for your needs.

What are 'Challenge Modes' in this skill?

After several rounds of questions, the skill shifts perspectives to act as a 'Contrarian' to test your logic or a 'Simplifier' to find a Minimum Viable Product (MVP) version of your idea.

Deep Interview

Name: Deep Interview
Author: Moliboy5000

byMoliboy5000

0•

生产力与工作流

Refines vague ideas into precise technical specifications using Socratic questioning and mathematical ambiguity scoring.

Deep Interview is a specialized capability for Claude Code designed to bridge the gap between abstract concepts and executable code. Inspired by the Ouroboros project, it replaces guesswork with a rigorous Socratic process that probes for hidden assumptions and measures clarity across dimensions like goals, constraints, and success criteria. The skill employs mathematical gating to ensure the AI truly understands the project scope, refusing to proceed to autonomous execution until ambiguity drops below a specific threshold. This ensures that the resulting plans and code align perfectly with user intent, eliminating the 'that's not what I meant' cycle in AI development.

主要功能

01Advanced 'Challenge Modes' including Contrarian and Simplifier perspectives

02Real-time mathematical ambiguity scoring with transparent progress reporting

03Socratic questioning loop targeting the weakest dimensions of project clarity

04Automated ontology extraction to track entity convergence and consistency

050 GitHub stars

06Integrated codebase exploration for brownfield project context mapping

使用场景

01Transforming high-level 'napkin sketch' ideas into detailed technical requirements

02Validating project assumptions before initiating autonomous execution pipelines

03Scoping complex modifications in existing codebases without manual documentation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add moliboy5000/.claude deep-interview

For use in Claude.ai and ChatGPT