KinCode — Requirements

Goals

• Help a family pick a strong, AI-clone-resistant shared safe word in under 10 minutes.
• Make the verification protocol — hang up, call back on a known number, ask the safe word — muscle memory for every adult in the circle.
• Surface each family member's voice-clone exposure so they understand why this matters specifically to them.
• Produce a one-page printable card the family can stick on a fridge or share to a parent's phone home screen.
• Keep all data on-device by default — a scam-defense app should never become its own data-leak risk.

Primary user

A 35–55-year-old "family CIO" — typically the adult child or caregiver who already manages shared subscriptions, holiday plans, and the parents' router. They have read at least one news story about AI voice cloning, are nervous about it, and want a single concrete thing they can do this weekend that protects everyone in the group chat. They are comfortable on a phone but will not download anything that requires an account, an SMS code, or a credit card on day one.

Functional requirements

• FR1: User can create a Trust Circle with a name and add 2–10 named members.
• FR2: Each member has a role tag of Anchor, Member, or Vulnerable (the latter triggers higher-risk default settings).
• FR3: The app suggests safe-word candidates from a curated wordlist that mixes nouns + adjectives and excludes obviously guessable phrases (pet names, sports teams, hometowns).
• FR4: A live strength meter scores the chosen safe word on length, uncommonness, and "would a scraper find this on social media" heuristics; explains the score in one sentence.
• FR5: User can lock the safe word; once locked, it is masked behind a click-to-reveal control with a configurable rotation reminder (default 6 months).
• FR6: For each member, the user can run a Voice Exposure self-audit (8–10 yes/no items) and see a 0–100 risk score with the top contributing factors.
• FR7: The Drill simulates an incoming AI-clone call with a timed transcript and at least three decision points (e.g. "Send the money", "Ask follow-up question", "Ask for the safe word", "Hang up and call back").
• FR8: Wrong drill answers route to a teaching post-mortem screen explaining what the scammer is exploiting; correct answers unlock the success state.
• FR9: A Print Card view renders the circle name, masked safe word (revealable), member list, and the three-step verification protocol on a single page sized for letter/A4.
• FR10: All state persists in localStorage so a refresh does not lose the circle; a one-click "Reset Demo" wipes it.
• FR11: All sample data is loaded from an inline JSON block so the page works when opened from file:// with no local server.
• FR12: The app is fully keyboard-navigable and has visible focus rings on every interactive control.

User stories

• As an adult child of aging parents, I want to set up a single shared safe word in under 10 minutes, so that everyone in my family knows what to ask if they get a frantic call.
• As a "family CIO", I want to see why my mom is high-risk for a voice-clone scam, so that I can have the conversation with concrete evidence rather than vague worry.
• As a college student with a public TikTok, I want a safe word I can memorise without writing it down, so that I can answer correctly even when I am genuinely panicked.
• As a caregiver running a drill with a parent, I want a guided simulated call, so that they experience the right behaviour once before they need it for real.
• As a financial advisor, I want a printable card I can leave with a client, so that the protocol survives between our quarterly meetings.
• As a privacy-conscious user, I want the safe word never to leave my browser, so that signing up for a scam-defense product does not itself become a leak.
• As a reseller (senior living community), I want to white-label the drill, so that I can offer it as a member benefit without building it.

Non-functional requirements

• Performance: page interactive in < 1 second on a mid-range phone; no blocking network requests at runtime.
• Privacy: zero outbound network calls in the prototype; all storage is localStorage; safe word is masked by default everywhere it appears.
• Accessibility: WCAG 2.2 AA color contrast; keyboard navigable; ARIA labels on the drill timer and the strength meter.
• Resilience: prototype must run from file:// (double-click index.html) without a local web server.
• Maintainability: vanilla HTML/CSS/JS, no build step, single CSS file, single JS file.

Out of scope (for the prototype)

• Real account creation, multi-device sync, or end-to-end encryption.
• Actual voice analysis, deepfake detection, or microphone access.
• Live integrations with FTC Sentinel, IC3, carrier APIs, or any backend.
• iOS/Android native wrappers and lock-screen widgets.
• Payments / subscription billing.

Open questions

• Should the safe word ever be exportable to an encrypted password manager (1Password, Bitwarden) — or is leaving the browser the wrong tradeoff?
• For the Vulnerable role, do we add a "trusted third party" (e.g. an advisor) to the circle, or keep the circle strictly family?
• How often should the rotation reminder default to — 6 months too aggressive, 12 too lax?
• Is the drill more effective as a one-shot ten-minute exercise, or as a 30-second monthly micro-drill that pings a member at random?