interview-simulator

name	interview-simulator
license	Apache-2.0
description	Designs and orchestrates a realistic interview simulation platform with voice AI, whiteboard evaluation, gaze-tracking proctoring, and mobile spaced repetition. Use for building mock interview infrastructure, configuring sessions, and adaptive difficulty. Activate on "interview simulator", "mock interview", "practice session", "voice mock". NOT for individual round-type coaching, resume writing, or prep timeline coordination.
allowed-tools	Read,Write,Edit,Bash,WebSearch,WebFetch
metadata	{"gated":true,"category":"Career & Interview","pairs-with":[{"skill":"interview-loop-strategist","reason":"Upstream orchestrator -- strategist decides WHAT to practice, simulator executes HOW"},{"skill":"senior-coding-interview","reason":"Provides coding round content, problem archetypes, and scoring rubrics"},{"skill":"ml-system-design-interview","reason":"Provides ML design problems, whiteboard evaluation criteria, 7-stage framework"},{"skill":"anthropic-technical-deep-dive","reason":"Provides Anthropic-specific technical topics and opinion evaluation"},{"skill":"tech-presentation-interview","reason":"Provides presentation structure, Q&A stress testing, and audience calibration"},{"skill":"values-behavioral-interview","reason":"Provides behavioral story bank, follow-up ladder depth, STAR-L format"},{"skill":"hiring-manager-deep-dive","reason":"Provides HM round structure, scope-of-impact evaluation, leadership signals"}],"tags":["interview","simulator","voice","whiteboard","practice"]}
category	Career & Interview
tags	["interview-prep","simulation","practice","mock-interviews","feedback"]

Interview Simulator

Platform architecture and coaching system for realistic mock interview practice with adaptive difficulty, emotion-sensitive voice AI, and comprehensive performance tracking.

Decision Points

Difficulty Adjustment During Session

IF emotion_data.hesitation_rate > 70% FOR 2+ consecutive responses
├── AND current_difficulty >= 3
│   └── REDUCE difficulty by 1 level
│       └── Voice persona: "Let's back up to fundamentals"
│
├── ELSE current_difficulty < 5
│   └── MAINTAIN difficulty
│       └── Voice persona: "Take your time, think through this"
│
└── ELSE (hesitation at difficulty 1-2)
    └── PAUSE session for confidence building
        └── Offer: "Would you like a 2-minute break?"

IF technical_accuracy > 85% AND response_time < 60% of target
├── AND current_difficulty < 5
│   └── INCREASE difficulty by 1 level
│       └── Add follow-up: "Now optimize for X constraint"
│
└── AND current_difficulty = 5
    └── MAINTAIN but add adversarial prompts
        └── "What if the requirements changed to..."

Round Type Selection

IF any dimension_score < 60 FOR last 3 sessions
├── SELECT lowest_scoring_round_type
│   └── Override user preference if different
│
ELIF no round_type practiced in last 7 days
├── SELECT oldest_unpracticed_round
│   └── Variety bonus for skills retention
│
ELSE
├── SELECT weakness_weighted_random()
│   └── 70% bias toward bottom 2 scores, 30% random

Proctor Flag Response

IF gaze_off_screen > 5 seconds
├── IMMEDIATE: Voice AI pauses, asks "Everything okay?"
│   └── IF happens again within 10 minutes
│       └── FLAG session as "distracted practice"
│       └── Debrief includes focus recommendations
│
IF second_monitor_detection = true
├── IMMEDIATE: Session pause, warning overlay
│   └── "Real interview won't have external resources"
│   └── Options: [Continue with flag] [Reset clean]
│
IF note_taking_detected (hand tracking)
├── ALLOW in Training mode with flag
├── BLOCK in Simulation mode
    └── "Please close notes for realistic practice"

Failure Modes

1. Overtraining Comfort Zone (Behavioral Looping)

Symptoms: >50% of sessions are same round type, plateau in target skill, false confidence in demo sessions Detection Rule: IF session_history.last_10_sessions contains >7 of same round_type THEN flag "comfort_zone_looping" Fix: Lock out preferred round type for 72 hours, force weakness-weighted selection, show heatmap of practice distribution

2. Emotion Detection False Positives (Technical Hesitation)

Symptoms: Difficulty drops during complex technical problems where thinking time is normal, candidate reports sessions feel too easy Detection Rule: IF difficulty_reduction triggered AND problem_complexity = high AND response_quality > 75% THEN flag "false_positive_hesitation" Calibrate hesitation baseline per round type, extend thinking time allowance for design problems from 30s to 90s, separate "thinking" from "struggling" in emotion model