# Mock Interview Room

> A mock job interview built from the job post you paste. An AI interviewer asks one question at a time, follows up when an answer has a gap, keeps a private note on each answer, and ends with a scored debrief per competency that quotes your own words. The interviewer can speak with an on-device voice, and you can answer by typing or speaking. No API key needed - each interviewer turn is paid with SkillSafe credits. https://mock-interview-room.skillsafe.ai/

## What it does

- Reads the job post in your browser: duties and requirements become a competency map (up to 6), plus behaviour competencies for the interview type. You can edit the map before starting.
- Optional résumé: questions get personal, and a free keyword match shows which job-post skills your résumé mentions and which it does not.
- Four interview types: mixed, behavioural (STAR stories), role skills, résumé walk-through. Entry, mid or senior level. 4, 6 or 8 questions.
- One question at a time. After each answer the interviewer writes a private note (strong, solid, thin or no answer, with a quote and one tip), hidden until you choose to see it, then asks the next question or one follow-up on a gap.
- Debrief: a verdict (strong hire to no hire), a 1-5 score per competency with a quote from your answers, why, and what to do next time; strengths, gaps against the job, a stronger outline for your weakest answer, and questions to practise next. Competencies that were never asked about score 0, not a guess.
- The browser checks the interviewer: quotes that are not in your answers, scores for competencies no question targeted, and an overall score that does not match the per-competency scores are all flagged.
- "Interview again on the gaps" starts a new interview on the same job aimed at your weaker competencies. Its debrief shows each competency's score last time and now, for every competency both interviews scored.

## Voice

- The interviewer's questions can be spoken by your browser's built-in voice, or by a natural voice (Kokoro-82M) that runs on your device over WebGPU after a one-time download. Nothing you hear is sent anywhere.
- Answers can be spoken with your browser's speech recognition (which may send audio to the browser's provider); the text lands in the answer box for you to edit before sending.

## Free vs paid

Free, no sign-in: the competency map, the keyword match, three recorded example interviews (data analyst, junior frontend developer, senior customer success manager), the voice, the transcript, and Markdown and JSON export. Paid: each interviewer turn (first question, each answer, the debrief), on credits, with the reserved amount shown before you send and the actual charge after. Signed-in interviews save to your SkillSafe account and resume on any device.

## How a turn is sent

There is no server-side chat session. Each turn is one stateless run carrying the job post and résumé (opening and closing parts kept, the middle replaced by a marker when long), the competency map, and a transcript window (the opening exchange plus the latest six lines). The debrief gets the whole transcript. The page tells you what was cut.

## API

https://mock-interview-room.skillsafe.ai/api.html - send `task` (open, answer, debrief) with job_post, resume, interview_type, level, competencies_json and the transcript; the reply is JSON Lines.