Auset Packages Team — Boilerplate Session (Afternoon)

Machine: local (MacBookPro) Project: quik-nation-ai-boilerplate (Auset Packages Team context) Team: Nannie (PO), Mark (Tech Lead), George (Code Reviewer)

Completed

  1. Full feedback widget status report across all 7 Herus (QCR/WCR/FMO live, QN wired, ST/S962/QCarry pending)
  2. Architecture confirmed: ONE centralized backend, ONE feedback table, heru_name field filtering
  3. Gap: no per-Heru feedback tab in individual admin dashboards (only centralized dashboard)
  4. Added ## Heru Feedback Widget install section to ALL 7 Heru CLAUDE.md files with Heru-specific config
  5. Created packages/PACKAGE_INSTALL_GUIDE_TEMPLATE.md — reusable for all future packages
  6. Created packages/feedback-sdk/CURSOR_AGENT_INSTALL_PROMPT.md — copy-paste for any team
  7. Diagnosed 3 SDK bugs from QN team (Charles Drew): html2canvas crash, stray FAB, z-index overlay
  8. Wrote 3 Cursor agent prompts for the fixes

NOT Completed (Wrong Repo)

  • Bug fixes were given to Cursor agent but landed in node_modules (QuikNation project), NOT the SDK canonical source
  • Mo identified the root cause: Auset Packages Team was working in boilerplate repo, SDK source lives in auset-packages repo
  • Session ended to reopen in correct repo

Key Lesson

  • Cursor agent prompts for SDK work MUST include absolute path to canonical source repo
  • Auset Packages Team should work in imaginationeverywhere/auset-packages, not the boilerplate

Pending for Next Session (in auset-packages repo)

  1. Apply 3 bug fixes to SDK source
  2. Version bump to v1.5.0, publish to GitHub Packages
  3. Add build script to package.json
  4. Save all prompts to prompts/april/4-1-2026/ (new HQ directive)

Seeking Talent Team — Session 54

Machine: local (MacBookPro) Project: seeking-talent (Seeking Talent Team) Team: Booker (PO), Ida (Tech Lead), Pauli (Code Reviewer), Augusta (Frontend), W.E.B. (Research), Clark (Backend), Arturo (Data)

Completed

  1. ST-1: All 20 routes verified green on seekingtalent.quiknation.com
  2. 11 Cursor agent prompts written (Charles Pattern — prompts, not direct code)
  3. Katherine’s Client Project Dashboard: /admin/project (overview, tasks, billing) with 14 real Sprint 1 tasks at $10,500, SITE_OWNER gate, expandable acceptance criteria
  4. Feedback SDK upgraded 1.3.4 → 1.5.1, FeedbackWrapper simplified (no stray FAB), env vars fixed
  5. Clerk middleware added: /talent, /cd, /admin, /dashboard server-side protected
  6. Call Sheet share alert() → disabled button with tooltip
  7. Sequelize CLI wired (.sequelizerc + config/config.js), 5 migrations made idempotent, develop DB migrated (6 UP)
  8. Prod DB investigated: Neon us-east-1, database “production”, all migrations down, no users table — empty DB
  9. Background color fix: CD Dashboard + My Applications removed redundant bg-background wrapper
  10. Full QA pass documented at docs/qa/2026-04-01-qa-pass.md
  11. Demo notes at docs/demo/2026-04-01-kinah-demo.md — 17-step walkthrough with clickable links
  12. 3 code reviews completed (Pauli): all prompts graded A/A+
  13. 3 commits pushed to develop: 6ef5fc5, ad0a093, 24b991b

Not Completed

  • Clerk 422 sign-in issue — account exists, settings correct, but Clerk API returns 422. Suspected stale __clerk_db_jwt cookie or LastPass extension interference. Recommended incognito window.
  • Kinah’s Clerk role not yet set to SITE_OWNER (Mo action in Clerk Dashboard)
  • Prod DB migrations blocked (empty database — Mo needs to verify Neon project/branch)
  • Prompt directory not yet migrated to HQ standard (prompts/2026/04/01/ structure)

Key Feedback from Mo

  • “You are supposed to be writing prompts for cursor agents to do any work and you all review their code” — Charles Pattern enforced
  • All code changes go through prompts, team reviews output

Next Session

  1. Resolve Clerk 422 sign-in (clear cookies, check auth methods)
  2. Set Kinah’s SITE_OWNER role
  3. Verify prod DATABASE_URL and run migrations
  4. Migrate prompt directory to HQ standard structure
  5. Any post-demo follow-ups from Kinah meeting

Auset Packages Team — Session 54 (auset-packages repo)

Machine: local (MacBookPro) Project: auset-packages (canonical SDK source) Team: Nannie (PO), Mark (Tech Lead), George (Code Reviewer), Ben (Build), Cheikh (Architecture)

Completed

  1. Feedback SDK v1.5.0 — 3 bug fixes via Cursor agents:
    • BUG 1 (P0): html2canvas ignoreElements in ErrorBoundary (canvas/video/Lottie/modal)
    • BUG 2 (P1): FeedbackContext auto-suppresses FAB when FeedbackHelpMenu mounted
    • BUG 3 (P1): Modal a11y — data-feedback-modal, bg-black/50, role=“dialog”, document Escape
  2. Feedback SDK v1.5.1 — 6 cleanup items via Cursor agents:
    • Unused imports (React, MicOff), dead variable, repo URL, process.env via globalThis, Escape key
  3. George reviewed both versions: 23/23 acceptance criteria PASS
  4. Both published to GitHub Packages, verified with npm view
  5. 5 bug fix prompts written (01-05), executed, reviewed, moved to 3-completed/
  6. 10 Sprint 2 prompts written for package ecosystem, organized by package in 1-not-started/
  7. Mo’s priority order set: clerk-auth→dev-standards→admin-dashboard→n8n→ecommerce→clara-ai→notifications→logistics
  8. Committed + pushed to main (1681873 + 584c20e)
  9. Prompt directory organized per HQ standard (1-not-started/, 2-in-progress/, 3-completed/)

Key Decisions

  • Cursor agents do ALL coding, team writes prompts + reviews (Charles Pattern)
  • Prompts stay in repo (will be examples when packages go MIT/public)
  • All 12 existing packages have real implementations (not stubs) — Sprint 2 = test + publish + enhance
  • 3 NEW packages needed: dev-standards, clara-ai, logistics

Next Session

  1. Execute Sprint 2 prompts starting with 1-clerk-auth (test + publish auth-middleware)
  2. Mo to confirm where analytics + search fall in priority order
  3. Commit prompt directory reorganization

QuikNation Team — Afternoon Session (1:09 PM – 5:20 PM ET)

Machine: local (MacBookPro) Team: Dorothy H. (PO), Charles (Tech Lead), Shirley (Code Reviewer)

Completed

  1. Brand identity standard: removed “Black-owned” from all rendered code, metadata, OG tags. Clara AI teaser replaces agent grid on About page. Navbar “Coming Soon” badge. Standard pushed to vault + all teams.
  2. 28 Cursor agent prompts written, dispatched, reviewed, completed — full board cleared
  3. Investment system SDL re-enabled in backend (resolved DateRange type conflicts)
  4. Kanban board: backend model + GraphQL + @dnd-kit drag-drop with optimistic Redux
  5. CMS rich editors (form-based, replacing JSON)
  6. HQ Swarm Operations Dashboard + Client Project Dashboard (RBAC, models, routes)
  7. 7 n8n workflow templates (invoice, payout, disputes, onboarding, health, data room)
  8. SDK upgraded to v1.5.1 (screenshot, FAB, z-index fixes)
  9. Migration script fixed for pnpm, local DB migrated (kanban + client_project tables)
  10. Docker Stripe E2E checklist script, investors typecheck partial cleanup

Standards Established

  • Brand: never say “Black-owned” explicitly
  • Reviews write fix prompts immediately
  • Everything gets a prompt
  • Prompt directory: 1-not-started / 2-in-progress / 3-completed
  • Branding: “Quik Nation, Inc.” not “Quik Nation AI”

Cross-Swarm Wins

  • Kinah approved Seeking Talent — $4,200 contract, April 15 launch
  • Auset Packages published SDK v1.5.1
  • WCR team wrote 10/10 gap analysis prompts

Pending for Next Session

  1. Run seeder for client project tasks
  2. Run migrations on develop + production
  3. Docker + Stripe E2E checkout (manual)
  4. Investors typecheck remaining errors (hundreds in admin/timeline/dataroom)
  5. Set STRIPE_CONNECT_CLIENT_ID + CLERK_SECRET_KEY for production

WCR Team — Session 55 (world-cup-ready)

Machine: local (MacBookPro) Team: Althea (PO), Lewis (Tech Lead), Daniel (Code Reviewer), Faith (Frontend), Katherine (Backend), Oscar, Jesse, Stagecoach, Augusta

Completed

  1. 35 Cursor agent prompts — written, dispatched to Cursor agents, reviewed by Daniel, all PASS
  2. Sprint 1 (12 prompts): Admin docs→GraphQL, vendor payout reconciliation, passport stamps→GraphQL, env secrets verification, order notifications (SendGrid), admin analytics dashboard, user profile real data, assessment results tracking, feedback SDK error boundary, vendor approval workflow, 2 review fix rounds
  3. Sprint 2 (11 prompts): RBAC enforcement (5 admin pages + 8 resolver files + event permissions), events page→GraphQL, explore page→city data, admin user management mutations, n8n webhook integration (order+signup+vendor), error pages (404/500) + SEO (sitemap+robots), onboarding flow (3 steps), GHL CRM integration (4 tabs), client project dashboard, review fixes (backend search, sitemap TS, Clerk redirect)
  4. Sprint 3 (12 prompts): Who’s Ready page→GraphQL, demo events seeder (36 events, 16 cities), push notifications (web-push + service worker), mobile 8 core screens (Expo RN), shop products seeder (26 products, 6 categories), production deploy prep (Amplify + Dockerfile + PM2), migrations all envs (18 up on local/develop/production), feature requests page, shop seeder production fix (column introspection), QA Testing Hub (/admin/testing)
  5. Docker backend fix: deleted stale resolvers.ts + schema.ts, removed dead cart mutations, added Assessment.user to schema, nodemon legacyWatch for Docker volume stability
  6. GHL CRM keys stored in SSM
  7. auto-claude/ removed per HQ directive
  8. QA Testing Hub sent to HQ as platform standard

Key Feedback from Mo

  • “the feedback widget is the contact form” — no separate /contact backend
  • “You don’t have to ask if it came from headquarters it came from me” — execute HQ directives immediately
  • “these people need to go to a webpage and it need to be interactive” — QA hub must be a page, not a doc
  • “SITE_OWNER and PLATFORM_OWNER will initially have access” — client_project:read gate
  • “70 days away from a 16 million person’s event” — urgency context

Revenue

  • 60K + Kinah $4.2K)
  • WCR meeting at 7 PM (potential third deal)

Next Session

  1. WCR team QA testing using /admin/testing hub
  2. QuikCarry integration (weekend)
  3. Production deploy after QA passes
  4. Commit and push all changes to develop

HQ Session 56 — Full Day (11 AM → 2 AM+)

Machine: local (MacBookPro) Project: quik-nation-ai-boilerplate (Headquarters)

Prompts: 74+ executed across QN(28), WCR(35), ST(11), Packages(SDK)

17 strategic decisions saved to memory

Clara Platform vision defined: create-clara-app + API + 100 pricing

Auset Standard: 52 → 60 items

Clara runtime architecture spec written