FULLIENCE · PRODUCT ACCEPTANCE · 2026-11-03

Every capability. Every boundary. One reviewable run.

Use isolated fixtures and Stripe sandbox. Current launch gates stay closed. This workbook tracks manual acceptance; it does not execute purchases, account changes or AI calls.

Read execution order and release criteria
# Fullience complete product QA sequence

Prepared 2026-11-03. This is a repeatable acceptance sequence, not a claim that every case has passed. All cases start Not run. Record the exact deployment, function versions, fixture and device for each execution.

## Run order and environments

1. Freeze baseline and confirm current gates. Do not open sales, make the site public, enable customer AI globally, or reset an owner account.
2. Run local deterministic regression checks and private marketing/mobile smoke.
3. Run authentication and role/access checks using isolated disposable accounts and an operator-controlled inbox.
4. Run connected agent stories with synthetic personal data and recorded action/provider usage.
5. Run payments and referrals in Stripe sandbox through the existing restricted test path. Never use real cards or change public billing flags to make a test pass.
6. Run negative ownership/failure cases on controlled fixtures.
7. Complete physical-device, accessibility and human credential/lifecycle cases.
8. Review P0 defects, reconcile billing/commissions and sign off the exact tested build.

Current customer AI-off and checkout-closed cases are release-blocker checks today. Enabled-workspace and payment success cases require isolated authorized fixtures; mark Blocked when unavailable rather than weakening launch controls. Each changed build requires affected regression cases to be rerun. Keep credentials, private resumes and session/payment tokens out of evidence. Use the product's supported reset flow only for explicitly disposable records. No test should contact employers, submit applications, send external outreach or execute live payouts.

## Golden customer stories

- Uncertain student: eligible signup → verification → AI-off dashboard; enabled fixture continues through project discovery → two direction hypotheses → small experiment → debrief → named roadmap.
- Targeted applicant: posting intake → fit/desirability review → six-document package → reviewed resume edit → interview practice/retry → application update → outcome debrief.
- Returning professional: saved goal/correction → weekly review → actual achievement → reviewed story/resume change → progression conversation.
- Paying sandbox customer: consent → eligible address → sandbox payment → authoritative webhook → allowance → same-task retry → renewal failure/recovery → verified cancellation.
- Partner: materials visible/personal URL hidden today → isolated attribution/payment fixture → reward/reversal → owner reconciliation.

## Pass criteria

A pass needs both the user-visible result and authoritative evidence where applicable: API status plus owned database record, sandbox event/ledger or persisted export. A prompt containing an instruction is not proof that the agent follows it. A file existing is not proof its content is correct. A screenshot is not proof billing is accurate. Record important negative results as well as successes.

P0 failures block release. P1 failures require disposition and an explicit accepted limitation before release. Blocked and Not run never count as Passed. No claim of legal compliance, hiring improvement or quantified conversion is established by this suite.

All cases begin Not run. Responsive browser checks do not replace real iOS/Android keyboard and touch tests. Keep personal data, credentials and card details out of notes.