deep-interview

SkillAI & models

Deep interview skill is a skill for AI agents that runs a Socratic interview, one question at a time, until a vague idea becomes a clear, approved plan. It scores ambiguity after each answer, checks the codebase for facts before asking, and only proceeds to execution once ambiguity drops below a threshold and the user explicitly approves.

Unavailable. Reviewed instructions are unavailable in ahel. This skill cannot be added until its reviewed copy is available.

Reviewed instructions are unavailable in ahel. This skill cannot be added until its reviewed copy is available.

What your AI can do with it

  • Asks one question per round, targeting the least clear aspect of the idea
  • Scores ambiguity after every answer to track how well-defined the plan is
  • Checks the codebase for facts before asking the user about them
  • Gates execution behind an ambiguity threshold and explicit user approval
  • Writes a specification file that hands off to a planning step

Signals

GitHub stars
39k
Forks
4k
Last commit
Oct 2026

Questions

How does the interview work?
It asks one question at a time, each round targeting the least clear aspect of your idea. After every answer it scores ambiguity, and it checks your codebase for facts before asking you about them.
When does it move to execution?
Only when ambiguity drops below a threshold and you explicitly approve. Until then it keeps interviewing.
What does it produce?
A specification file that hands off to a planning step.
Advanced
Item type
skill
Key
deep-interview-yeachan-heo
Source
github.com/yeachan-heo/oh-my-claudecode