Expected Output Interview Template
Paste this prompt into a fresh Claude Code session to start the interview for any tier.
Before starting, replace the [PLACEHOLDERS] with the specific tier you're building the spec for.
Prompt
Read these files first (in order):
1. `C:\dev\knowledge\processes\expected-output-methodology.md` — the methodology
2. `C:\dev\knowledge\products\chatbot\tier-restrictions.md` — the feature matrix
3. `C:\dev\knowledge\data\features.yaml` — what each tier includes
4. `C:\dev\knowledge\processes\customer-facing\onboarding.md` — the onboarding flow
5. `C:\dev\knowledge\processes\customer-facing\checkout-flow.md` — checkout details
6. `C:\dev\churchwiseai-web\src\lib\tier-config.ts` — TIER_FEATURES code
7. `C:\dev\churchwiseai-web\src\lib\tool-config.ts` — tool tier assignments
8. `C:\dev\churchwiseai-web\src\lib\agent-type-config.ts` — TIER_AGENTS
Then read the current state of a [TIER] customer from the database:
- DB: `SELECT * FROM premium_churches WHERE [APPROPRIATE_FILTER]`
- DB: `SELECT * FROM church_voice_agents WHERE [APPROPRIATE_FILTER]`
Now, we are going to build the **[TIER] Expected Output Spec** together.
This is an INTERVIEW. You will walk me through every touchpoint in the methodology
(Phases 2-3), one at a time. For each touchpoint:
1. Tell me what touchpoint we're defining (e.g., "Touchpoint 18: Dashboard Header")
2. Show me what currently exists (WebFetch the production page or read the component code)
3. Ask me: "Is this correct for [TIER]? What should be visible? What should be hidden?"
4. Capture my answer in the spec
Start with the user state definition, then go through touchpoints A (Pre-Purchase),
B (Emails), C (Public Pages), D (Dashboard — every tab and sub-tab), and
E (Lifecycle Events) in order.
Do NOT skip touchpoints. Do NOT assume you know the answer. ASK ME.
Save the spec to `C:\dev\knowledge\acceptance\[TIER_FILENAME].md` as we go,
updating it after each section.
After all touchpoints are covered, generate the Playwright E2E test from the
approved spec and save it to
`C:\dev\churchwiseai-web\e2e\journeys\[TIER_FILENAME]-customer.spec.ts`.
This is the most important document in the entire codebase. Take your time.
Tier-Specific Values
| Tier | Filename | DB Filter | Notes |
|---|---|---|---|
| Starter Chat | starter-chat | plan = 'starter' AND channel = 'chat' | Most constrained — build this first |
| Starter Voice | starter-voice | plan = 'starter' AND channel = 'voice' | |
| Starter Both | starter-both | plan = 'starter' AND channel = 'both' | |
| Pro Chat | pro-chat | plan = 'pro' AND channel = 'chat' | |
| Pro Both | pro-both | plan = 'pro' AND channel = 'both' | Most popular expected combo |
| Suite Chat | suite-chat | plan = 'suite' AND channel = 'chat' | |
| Suite Both | suite-both | plan = 'suite' AND channel = 'both' | Full feature set |
| Pro Website | pro-website | plan = 'pro_website' | PewSearch-specific |
| Free Claim | free-claim | plan IS NULL (recently claimed) | Minimal chatbot |
| Trial Expired | trial-expired | status = 'preview' AND trial expired | Cross-tier |
| Cancelled | cancelled | status = 'cancelled' | Cross-tier |
Recommended Order
- Starter Chat — most constrained tier, most likely to have bugs. Build first.
- Pro Both — most popular expected plan. Many answers will be "same as Starter except..."
- Suite Both — full features, mostly additive from Pro.
- Pro Website — different product, different touchpoints.
- Remaining tiers — differential from the ones already defined.
- Trial Expired / Cancelled — lifecycle states, cross-tier.
First Interview: Starter Chat
For the very first spec (Starter Chat), use this exact prompt with Zewdei's test church:
Read these files first (in order):
1. `C:\dev\knowledge\processes\expected-output-methodology.md` — the methodology
2. `C:\dev\knowledge\products\chatbot\tier-restrictions.md` — the feature matrix
3. `C:\dev\knowledge\data\features.yaml` — what Starter includes
4. `C:\dev\knowledge\processes\customer-facing\onboarding.md` — the onboarding flow
5. `C:\dev\knowledge\processes\customer-facing\checkout-flow.md` — checkout details
6. `C:\dev\churchwiseai-web\src\lib\tier-config.ts` — TIER_FEATURES code
7. `C:\dev\churchwiseai-web\src\lib\tool-config.ts` — tool tier assignments
8. `C:\dev\churchwiseai-web\src\lib\agent-type-config.ts` — TIER_AGENTS
Then read the current state of Zewdei's church (our live Starter Chat customer):
- DB: `SELECT * FROM premium_churches WHERE church_id = '96f5b89e-b238-4811-8d76-777f84f4241c'`
- DB: `SELECT * FROM church_voice_agents WHERE church_id = '96f5b89e-b238-4811-8d76-777f84f4241c'`
Now, we are going to build the **Starter Chat Expected Output Spec** together.
This is an INTERVIEW. You will walk me through every touchpoint in the methodology
(Phases 2-3), one at a time. For each touchpoint:
1. Tell me what touchpoint we're defining (e.g., "Touchpoint 18: Dashboard Header")
2. Show me what currently exists (WebFetch the production page or read the component code)
3. Ask me: "Is this correct for Starter Chat? What should be visible? What should be hidden?"
4. Capture my answer in the spec
Start with the user state definition, then go through touchpoints A (Pre-Purchase),
B (Emails), C (Public Pages), D (Dashboard — every tab and sub-tab), and
E (Lifecycle Events) in order.
Do NOT skip touchpoints. Do NOT assume you know the answer. ASK ME.
Save the spec to `C:\dev\knowledge\acceptance\starter-chat.md` as we go,
updating it after each section.
After all touchpoints are covered, generate the Playwright E2E test from the
approved spec and save it to
`C:\dev\churchwiseai-web\e2e\journeys\starter-chat-customer.spec.ts`.
This is the most important document in the entire codebase. Take your time.