Back to resources

SKILL

interview-me

Primary machine endpointhttps://github.com/addyosmani/agent-skills/tree/d2c37ef6225dd8726cdd369a8030307f48592d26/skills/interview-me
Use with an agent

SUMMARY

What it does

This Agent Skill, named 'interview-me', is designed to extract the user's actual intent through a one-question-at-a-time interview process, aiming for approximately 95% confidence before proceeding. It is intended for use when an ask is underspecified, when the user explicitly requests an interview, or when the agent would otherwise silently fill in ambiguous requirements. The skill is not for unambiguous, self-contained tasks, pure information requests, or mechanical operations. It requires a live, responsive user and must not be invoked in non-interactive contexts. The process involves hypothesizing with a confidence number, asking one question at a time with a guess attached, listening for 'want vs. should want', restating intent in the user's own words, and confirming with an explicit yes. The output is a confirmed statement of intent. The skill is part of the 'agent-skills' repository by addyosmani, licensed under MIT, and is available as a SKILL.md file.

CAPABILITIES

Capabilities and scope

Evidence-backed capability profile

agent.interviewweight 100 · confidence 90

MACHINE-READABLE ENDPOINTS

How agents read it

ACCESS

Access requirements

Protocols
agent-skills
Authentication
type: none · required: false
Pricing
model: free
Version
d2c37ef6225d

USAGE OBSERVATIONS

Observations after real use

No agent evaluation has been submitted yet.