Skip to content

Latest commit

 

History

History
92 lines (66 loc) · 4.02 KB

File metadata and controls

92 lines (66 loc) · 4.02 KB

Controlled Beta Guide

Current status

This product is ready for self-testing, 5–10 trusted student/friend testers, and counselor feedback after the QA checklist passes. It is not ready for full education-center pilots, paid institution use, or production Firebase dashboards.

Who can test now

  • The project owner/founder.
  • Trusted international students or recent applicants who understand this is under validation.
  • Parents/friends using synthetic or safely redacted messages.
  • Counselors/advisers reviewing language, workflows, and escalation guidance.

Who should not use it yet

  • Schools or education centers making operational decisions.
  • Paid institution customers.
  • Minors testing alone without a trusted adult/counselor.
  • Anyone facing immediate financial loss, threats, self-harm risk, immigration deadlines, or account compromise without human support.

What testers must be told

  • The tool detects risk indicators, not certainty.
  • It is educational and under validation.
  • It does not provide legal, immigration, financial, emergency, or law-enforcement advice.
  • Checker analysis runs locally by default.
  • Redacted reporting is best effort and requires preview review.
  • localStorage reports are browser-only, not encrypted, and not institutional records.
  • Firebase dashboard mode is not connected unless a secure Firebase implementation and governance review are completed.

What data testers must not paste

Do not paste:

  • Passport scans or passport numbers.
  • Student IDs, national IDs, visa numbers, SEVIS/CAS/I-20 identifiers.
  • Card numbers, bank details, CVV, PINs, passwords, OTPs, recovery codes.
  • Exact addresses, phone numbers, personal email addresses, names of minors.
  • Private documents, screenshots with personal data, or legal/immigration files.

Safe testing with synthetic messages

  • Start with built-in demo messages.
  • Create synthetic examples by scenario: scholarship fee, visa threat, fake housing deposit, bank/card phishing, testing score upgrade.
  • Do not use real names, exact institutions, or real account numbers.
  • Record only category, risk level, whether the explanation made sense, and next-step clarity.

Safe testing with real messages after redaction

  • Remove names, IDs, exact URLs, account numbers, addresses, phone numbers, and institution-specific identifiers before testing.
  • Prefer summarizing the scenario instead of pasting full text.
  • Use the report page only after reviewing the redacted preview.
  • If redaction misses something, stop and clear the form/local data.

How to collect feedback without raw messages

  • Use the anonymous feedback controls in the result card.
  • Use docs/FEEDBACK_ANALYSIS_TEMPLATE.md for sessions.
  • Store categories and structured observations only.
  • Never paste raw messages into docs, issues, pull requests, emails, spreadsheets, or chat.

When to stop testing

Stop immediately if:

  • A tester pastes highly sensitive data.
  • A tester is scared, pressured, or confused about a real deadline.
  • A result appears to encourage unsafe action.
  • A legitimate message scores high/critical and the user may ignore an important official deadline.
  • A high-risk message scores low and the tester might proceed unsafely.
  • localStorage or Firebase behavior is unclear.

When to involve a counselor/adult

Involve a counselor, trusted adult, bank, official school office, or relevant authority when:

  • Money, deposits, tuition, refunds, cards, bank accounts, passwords, OTPs, or identity documents are involved.
  • A student is a minor.
  • The message threatens visa cancellation, deportation, admission loss, account closure, or housing loss.
  • The student already clicked, paid, or shared data.
  • The student feels afraid or pressured.

How to summarize feedback safely

Use broad categories:

  • Scenario category, not message text.
  • Expected/actual risk level.
  • Whether next steps were understood.
  • Whether wording felt too scary, too certain, or unclear.
  • False-positive/missed-risk notes without identifiers.
  • Privacy concerns and feature requests.