Skip to main content

Session Handoff: Persona Testing & Product Hardening

For the next agent session: Read this entire document, then execute the phases in order. Use sonnet model for all subagents unless noted otherwise. The CEO tested Starter Chat yesterday and found 40+ issues. Two overnight agents created a retrospective and pre-audited other products. This session finishes the work.

Phase 1: Merge Overnight Hardening Branches (10 min)​

Three feature branches have fixes from the overnight hardening audit. Review and merge each.

Task 1.1: Review and merge feat/harden-itw​

  • Codebase: C:\dev\sermon-illustrations
  • Changes: Forces currency: 'usd' on Stripe checkout
  • Action: cd /c/dev/sermon-illustrations && git diff master feat/harden-itw — review, then merge to master and push
  • Model: sonnet

Task 1.2: Review and merge feat/harden-pewsearch​

  • Codebase: C:\dev\pewsearch\web
  • Changes: Adds plain-text URL fallback to welcome email
  • Action: cd /c/dev/pewsearch/web && git diff master feat/harden-pewsearch — review, then merge to master and push
  • Model: sonnet

Task 1.3: Review and merge feat/harden-sermonwise​

  • Codebase: C:\dev\churchwiseai-web
  • Changes: Fixes duplicate embed widget in Starter features, role count 7→9, agent count 2→4 for Pro, illustration count 30K→50K
  • Action: cd /c/dev/churchwiseai-web && git diff main feat/harden-sermonwise — review, then merge to main and push
  • Model: sonnet

Task 1.4: Add adaptive_pricing: { enabled: false } to remaining checkout routes​

The overnight audit flagged this as not-yet-fixed for:

  • sermon-illustrations/src/app/api/stripe/checkout/route.ts
  • pewsearch/web/src/app/api/stripe/checkout/route.ts
  • pewsearch/web/src/app/api/stripe/pre-checkout/route.ts

CWA's checkout-embedded already has it. Add to the other three routes.

  • Model: sonnet

Phase 2: TAG Consistency Sweep (30 min)​

Use the TAG registry at knowledge/tests/tag-registry.yaml to verify every tagged claim is consistent across all codebases. Dispatch one agent per TAG group.

Task 2.1: Verify #tools_count​

  • Canonical: Starter=12, Pro=35, Suite=39
  • Search all 3 codebases for any reference to tool counts (12, 33, 35, 39, "ministry tools")
  • Fix any inconsistencies (the overnight audit found some in StripeSetupTab)
  • Model: sonnet

Task 2.2: Verify #agent_count​

  • Canonical: Starter=2 (Care + Coordinator), Pro/Suite=4 (+ Discipleship + Stewardship)
  • Search for: "2 agents", "4 agents", "care agents", agent lists
  • Cross-check: pricing pages, demo pages, product pages, emails, Stripe product descriptions, starter kit PDF references
  • Model: sonnet

Task 2.3: Verify #pricing across all surfaces​

  • Check: Every price on every marketing page matches PRICING.md
  • Check: Stripe live product descriptions match marketing
  • Check: Emails reference correct prices per tier
  • Check: Trial mentions are accurate (14 days chat only, no trial voice/bundle)
  • Model: sonnet

Task 2.4: Verify #church_count​

  • Canonical: 218,000+ (directory_visible=true)
  • Search for: "218", "240", "260", "261" across all codebases
  • Fix any that say 240K or 261K
  • Model: sonnet

Task 2.5: Verify #demo_numbers​

  • Canonical: US (469) 615-2221, CA (365) 825-4095
  • NOT: (888) 603-0316 (sales line, not demo)
  • Search all codebases + PDF for phone numbers
  • Model: sonnet

Task 2.6: Jargon scan (#jargon_forbidden)​

  • Scan all dashboard components (churchwiseai-web/src/app/admin/, churchwiseai-web/src/components/admin/)
  • Flag any forbidden terms from the registry that appear in user-facing labels without tooltips
  • List findings but don't fix (FA-016 covers the full label pass)
  • Model: sonnet

Phase 3: Persona Testing — CWA Products (1.5 hours)​

Run 3 persona agents against each product the CEO plans to test next. Each persona walks the full journey and reports findings. Use knowledge/tests/persona-test-prompts.md for the persona prompts.

Task 3.1: PewSearch Premium ($9.95/mo) — 3 personas​

Dispatch 3 agents IN PARALLEL (all sonnet):

Agent A — Tired Pastor:

[Insert Tired Pastor prompt from persona-test-prompts.md]

Your task: Walk through the PewSearch Premium signup journey.
1. Go to pewsearch.com
2. Find a church (search for a real church name)
3. Find the "Claim this church" flow
4. Walk through the claim form
5. Reach the checkout page (DO NOT pay)
6. Note every moment of confusion, every unclear label, every missing explanation
7. Check the pricing page at pewsearch.com/pricing

Cross-reference the TAG registry at knowledge/tests/tag-registry.yaml.
Report all findings as a markdown list with severity (critical/important/minor).

Agent B — Justice-Minded Fact Checker:

[Insert Justice-Minded prompt]

Your task: Verify every claim on pewsearch.com.
1. Count features listed for Premium vs Pro Website
2. Verify church count (should be 218K+, NOT 240K or 261K)
3. Check that pricing matches PRICING.md
4. Read every FAQ answer for accuracy
5. Check demo page claims
6. Verify the claim flow doesn't create DB records before payment

Report all findings.

Agent C — Anxious Board Member:

[Insert Anxious Board Member prompt]

Your task: Evaluate PewSearch from a risk/safety perspective.
1. Read the privacy policy — is it reassuring?
2. Read the terms of service — any red flags?
3. What happens to church data if they cancel?
4. Is there a clear "off switch" for the chatbot?
5. How is the AI disclosed to visitors?

Report all findings.

Task 3.2: IllustrateTheWord Premium ($9.95/mo) — 3 personas​

Dispatch 3 agents IN PARALLEL (all sonnet):

Agent A — Tired Pastor:

Walk through illustratetheword.com.
1. Browse free illustrations
2. Find the signup/premium CTA
3. Is the value proposition clear? Why pay $9.95?
4. Try to sign up (Supabase Auth)
5. Find the upgrade path
6. Note every moment of confusion

Agent B — Justice-Minded Fact Checker:

1. Count illustrations — does the site say 30K+? Verify.
2. Check pricing — is $9.95 clear?
3. Is there a free trial mentioned? (There shouldn't be for ITW)
4. Are the illustration categories accurate?
5. Check if the checkout forces USD

Agent C — Budget Treasurer:

1. Is $9.95/mo clear with no hidden fees?
2. Can you tell what you get vs free?
3. Is there pressure to upgrade?
4. Is cancellation obvious?

Task 3.3: SermonWise Pro ($19.95/mo) — 3 personas​

Dispatch 3 agents IN PARALLEL (all sonnet):

Agent A — Tired Pastor:

Go to sermonwise.ai.
1. Does it load? (hostname rewrite to churchwiseai.com/sermons)
2. Is the product clear — what does SermonWise do?
3. Find pricing/signup
4. Walk through signup flow
5. Note confusion points

Agent B — Justice-Minded Fact Checker:

1. Verify pricing ($19.95/mo)
2. Count features listed — are they all real?
3. Check theological traditions claim (17)
4. Is there a free tier? What does it include?
5. Verify checkout forces USD

Agent C — Catholic Secretary:

1. Does SermonWise support Catholic homily preparation?
2. Is "sermon" the only term used? (Should offer "homily" for Catholics)
3. Are Catholic-specific illustrations available?
4. Does the theological lens for Catholic actually work?

Task 3.4: CWA Pro Chat Dashboard ($34.95/mo) — 2 personas​

Use the existing demo church (Grace Community) or create a test Pro account.

Agent A — Tired Pastor (Pro tier):

Access the CWA admin dashboard for a Pro Chat church.
1. Do all 4 agents show? (Care, Coordinator, Discipleship, Stewardship)
2. Are Pro features unlocked? (FAQ management, document upload, analytics)
3. Are voice features hidden? (chat-only plan)
4. Does the training simulator work?
5. Can you add FAQs?
6. Can you upload a document?
7. Does the analytics dashboard load?

Agent B — Justice-Minded Fact Checker (Pro tier):

1. Count tools available — should be 35 for Pro
2. Verify 4 agents are functional, not just displayed
3. Check that Pro Website is mentioned as included
4. Verify the upgrade comparison shows Suite correctly
5. Check that embed widget customization works

Phase 4: Compile Results & Create Action Items (30 min)​

Task 4.1: Merge all persona reports​

  • Combine findings from all persona agents into a single document
  • Categorize by severity: Critical (blocks testing), Important (CEO will notice), Minor (polish)
  • Save to knowledge/tests/persona-test-results-2026-03-31.md

Task 4.2: Fix critical issues​

  • Any critical findings should be fixed immediately on feature branches
  • Important findings logged as backlog items in FOUNDER_ACTIONS.md

Task 4.3: Update TAG registry​

  • If any TAG canonical values were wrong, update knowledge/tests/tag-registry.yaml

Task 4.4: Update decision log​

  • Append results to C:\dev\DECISION_LOG.md

Execution Notes​

  • All subagents use sonnet model unless the task requires architectural judgment (use opus)
  • Parallel dispatch: Tasks within a phase that are independent should be dispatched in parallel
  • Sequential phases: Phase 1 → 2 → 3 → 4 (each depends on the previous)
  • Do NOT push to main/master without reviewing changes first
  • Do NOT write test data to production Supabase
  • Use WebFetch for testing live URLs (the products are deployed on Vercel)
  • Read each product's CLAUDE.md before working in that codebase

Key Files to Read First​

FileWhy
C:\dev\CLAUDE.mdCross-project rules
C:\dev\PRICING.mdSource of truth for all prices
knowledge/tests/tag-registry.yamlTAG consistency registry
knowledge/tests/persona-test-prompts.mdPersona prompt library
knowledge/processes/manual-testing-retrospective-2026-03-30.mdWhat went wrong yesterday
knowledge/tests/pre-ceo-hardening-audit-2026-03-30.mdWhat the overnight audit found