Claude Code skill
User interviews
Open source

simulated-interview

Claude plays a persona in a jobs-to-be-done interview, and every insight is marked synthetic.

A model playing a user drifts toward agreeing with you. Here the persona is told to push back, answer from past behaviour and say “I don’t know”, so weak assumptions show up before you spend time on real people.

Year: 2026
Needs: Claude Code
Optional tools: Sub-agents
License: MIT

View on GitHub

Install

git clone https://github.com/VladPolyakov-art/simulated-interview ~/.claude/skills/simulated-interview

What it writes

Fixed file names, so the other skills in the kit find them without asking. Each file also gets a row in research/INDEX.md.

research/interview-<persona-slug>.md: The debrief: what held up, new insights, edge cases, contradictions, questions for real research.
research/interview-<persona-slug>/transcript.md: The full dialogue, every persona turn tagged synthetic.
research/interview-<persona-slug>/guide.md: The question guide, if you want to keep it.

How it works

Two modesLive: you ask and Claude answers as the persona. Guided: Claude writes the guide and plays both sides.
A persona that pushes backLeading questions get a real answer. No statistics, prices or quotes from “other users”.
Synthetic stays syntheticNo “users want”, no counts or percentages in the debrief. The sample is one simulated person.

What the agent does

01ReconnaissanceLooks for personas, voice of users and audits in research/.
02IntakeThe persona, what we are exploring, the riskiest assumptions, the mode.
03Persona sheetEvery line keeps its source tag. What Claude adds is marked as an assumption.
04Question guideWarm-up, past behaviour, pain, then the idea.
05InterviewClaude stays in role until you say “exit role”.
06DebriefSix sections, each insight tied to a turn in the transcript.

Works with the kit

None of these is required. With an empty research/ folder it works from what you give it, and when a sibling’s file is there, it reuses it instead of redoing the work.

Reads

proto-personasThe persona, the primary one by default.
competitive-researchRealistic pains and the tools the persona uses today.
ux-auditConcrete friction for past-behaviour questions.
deep-researchGaps and open questions, as candidate risky assumptions.

Read by

feature-researchHypotheses and questions for its user section, still synthetic.
proto-personasSuggested updates to the persona.
deep-researchThe questions for real research.
ux-auditA persona’s expectations, as hints for what to look at.

Part of the Product research kit: six skills, one research/ folder and one set of evidence tags.