Audit · Things Started But Not Finished

Date: May 19, 2026 Method: Cross-referenced Fly secrets, codebase modules with env-gated init, conversation history, and prior promises.

Everything below is something you started, paid for, mentioned, or promised, but isn't producing daily value yet.

---

🔴 RED · Critical for launch, not finished

1. Sentry (your team is asking)

Sign up here: https://sentry.io/signup Free tier: 5K errors/month, 10K performance events. Plenty for launch. Upgrade later if needed.

Setup (5 min):

1. Sign up with the email you want as org admin (probably brian@krytera.com or brian@posture.vision)
2. Org name suggestion: krytera or posturevision
3. Create 4 projects (one per service):
   • pv-auth-service (Node.js backend)
   • posture-vision-frontend (React)
   • lead-seg-worker (Node.js worker on Fly)
   • posturevision-sitemap (browser/Cloudflare Pages)
4. Sentry gives you a DSN per project (looks like https://abc123@o123456.ingest.sentry.io/4567)
5. Set as env vars on each service:
   • Fly worker: flyctl secrets set SENTRY_DSN=<dsn> -a lead-seg-worker
   • Vercel pv-auth-service: add SENTRY_DSN env var
   • Frontend: add VITE_SENTRY_DSN (or NEXT_PUBLIC_SENTRY_DSN)
6. Install SDK in each service (npm install @sentry/node or @sentry/react)
7. Add 3 lines to each app's entry point to init Sentry with the DSN

Want me to wire the Fly worker side? Say "wire Sentry" and I'll add init code + a test error endpoint. ~15 min.

Invite teammates after setup: Settings → Members → Invite → christiane@krytera.com, alain@..., etc.

2. Stripe Foundation Cohort price routing (still unverified)

Status: You said it's wired live in GHL. Hasn't been tested with a real cohort tag. Action: Apply foundation-cohort-2026 tag to your own contact, click the signup link, confirm checkout shows $9.99/mo correctly. Test today. Risk if not done: May 24 outreach sends patients to a broken checkout.

3. GHL Workflow 1 — 22 actions

Status: Trigger saved. 22 follow-up actions not built. Action: Open the workflow in Brave + paste the prompt from /docs/paste-blocks.html Block 1 OR follow /docs/ghl-workflows.html manually. Risk if not done: Foundation Cohort outreach doesn't fire May 24.

4. GHL AI Agent system prompt

Status: Spec written. Not pasted into GHL. Action: GHL PV subaccount → Conversational AI → System Prompt → paste from /docs/voice-agent-script.html (the full system prompt block at the bottom). Risk if not done: Inbound DMs during launch week land on you personally, no triage.

5. 300 patient bulk tag

Status: Not applied. Action: GHL Contacts → filter patient list → Bulk Actions → Add Tag → foundation-cohort-2026. Must complete before May 24. Risk if not done: Workflow has no audience to fire to.

---

🟡 YELLOW · Started, paid, not producing daily value yet

6. ElevenLabs (✅ key set, ❓ voice clone status)

Fly secret present: ELEVENLABS_API_KEY ✅ Status unclear: Have you actually cloned your voice yet? The engine has voice synth code wired but I don't see evidence of a saved voice ID. Action: Log in at https://elevenlabs.io/app, go to Voices → Create a voice → Instant Clone → upload ~1 min of clean audio of you talking → save the voice ID Then: tell me the voice ID and I'll set ELEVENLABS_VOICE_ID_BRIAN on Fly. Engine can then voice-over screen recordings + AI Elina videos.

7. HeyGen (not configured, code is waiting)

Fly secret NOT set: HEYGEN_API_KEY is missing Code exists: lib/heygen.mjs and jobs/video-generator.mjs are written and waiting for the key Setup doc: Code/lead-seg-worker/HEYGEN_SETUP.md walks you through it Cost: $24/mo Creator plan + ~10 min to pick 3 avatars What you get when wired: engine takes TikTok scripts from overnight cron + auto-renders them as AI Elina videos. Zero filming for video content.

8. Imagen 3 (Vertex AI) and Gemini Nano Banana (not configured)

Fly secrets NOT set:

• GOOGLE_VERTEX_PROJECT ❌
• GOOGLE_APPLICATION_CREDENTIALS_JSON ❌
• GOOGLE_AI_STUDIO_KEY ❌

Code exists in lib/ai-image.mjs: the AI image fallback ladder will try Imagen → DALL-E → Gemini in order, but right now only DALL-E is wired. So images you generate via the engine only fall back to DALL-E (~$0.04/img), never the cheaper Gemini ($free tier) or higher-quality medical-grade Imagen ($0.02/img).

Recommendation: add Gemini in 5 min (free, lowest risk), skip Imagen until you have specific need for medical-grade imagery.

Gemini setup (5 min):

1. https://aistudio.google.com/app/apikey → Create API key (uses your existing Google account)
2. flyctl secrets set GOOGLE_AI_STUDIO_KEY=<key> -a lead-seg-worker
3. Done. Engine starts using it as fallback.

9. PostHog (you just invited the team ✅)

Status: ✅ Confirmed by Brian. Alain and Christiane invited. Next: verify they accept the invite. Once they're in, walk them through:

• Where to find session replays (for understanding scan-flow drop-off)
• Funnel for Signup → First scan → Day-7 retention
• The dashboard you've built (if any)

10. Slack alert webhook (live but underused)

Fly secret present: SLACK_ALERT_WEBHOOK_URL ✅ Current usage: Worker health check failures fire here every 15 min if engine breaks Underused for: Launch-day alerts, conversion notifications, brand-mention high-value hits (currently fire Pushover, not Slack) Recommendation: add Slack mirrors for the same events so the team sees them, not just you. ~30 min wiring.

11. Pushover (working ✅)

Status: ✅ Live, brand mentions + conversions + slot decisions all fire here. Underused for: none, this is healthy.

12. Tavily (wired into 2 jobs but could do more)

Fly secret present: TAVILY_API_KEY ✅ Currently powers: trending scraper + brand mention monitor Could also power: competitive intelligence (Squat U / Aaron Alexander / Foundation Training new content alerts), press monitoring beyond just "PostureVision" mentions

13. Meta Graph API (for both subaccounts)

Fly secrets present: META_GRAPH_TOKEN_DH, META_GRAPH_TOKEN_PV, META_IG_USER_ID_DH, META_IG_USER_ID_PV ✅ Currently powers: engagement ingestion every 14:00 UTC Token expiration risk: Meta tokens typically expire every 60 days. If you set these in March/April, they may expire mid-launch. Test now: hit your IG insights endpoint manually and confirm it doesn't return 401.

14. Resend (transactional email)

Fly secret present: RESEND_API_KEY ✅ Currently powers: unclear — needs verification Action: confirm Resend is the email provider for transactional emails (welcome, receipt, password reset) OR confirm GHL handles all email. If both are wired, decide which is canonical.

---

🟢 GREEN · Started and producing daily value

• ✅ Anthropic Claude API (powers all content generation)
• ✅ Bannerbear (template renders working)
• ✅ Unsplash (stock photos on Generate flow)
• ✅ OpenAI / DALL-E (AI image fallback)
• ✅ Pushover (all critical alerts)
• ✅ Notion (13 databases all wired)
• ✅ Cloudflare Pages (2 sites: sitemap + operating docs)
• ✅ Fly worker (autonomous engine, 9 cron stages)
• ✅ GHL API (both DH and PV subaccounts)
• ✅ Tavily (research + brand mentions)
• ✅ Slack alert webhook (health-check alerts)
• ✅ Meta Graph (engagement ingestion)
• ✅ PostHog (analytics ingestion, just invited team)

---

💸 Subscriptions / services you've mentioned but I can't verify

I have no way to check these from inside the engine. Tell me which ones you actually have active:

Service	Why I think you might have it	Status (you fill in)
CapCut Pro ($7.99/mo)	I recommended for video editing	❓ subscribed?
Opus Clip ($19/mo)	I recommended for long → short cuts	❓ subscribed?
HeyGen ($24/mo Creator)	For AI Elina video pipeline	❓ subscribed?
Sentry (free tier)	Team is asking for it now	❌ not yet (see above)
Shutterstock	You mentioned having an account	✅ confirmed
Cloudflare Access	Gates your sitemap + content engine	✅ confirmed (need to add Christiane)
Stripe (via GHL)	Live payment processing	✅ confirmed live
GHL (both subaccounts)	Calendar + CRM + AI agents	✅ confirmed live
Notion	All operating data	✅ confirmed live
Fly.io	Worker hosting	✅ confirmed live ($5/mo machine)
Cloudflare Pages	Free tier for both sites	✅ confirmed live
Vercel	pv-auth-service hosting	❓ confirm
OpenAI ($0.04/img DALL-E)	AI image fallback	✅ confirmed (key set)

---

🎯 What I'd ship today to clear the runway

In order of effort × launch impact:

1. Sign up for Sentry (5 min) → unblocks team request, gives error visibility for launch week
2. Set up Gemini AI Studio key (5 min) → strengthens AI image fallback ladder for free
3. Apply 300-patient bulk tag in GHL (10 min) → makes May 24 outreach actually fire
4. Verify Stripe Foundation Cohort routing (5 min) → highest-risk pre-launch check
5. Paste GHL AI Agent system prompt (15 min) → DM avalanche protection
6. Verify ElevenLabs voice clone exists + grab voice ID (10 min) → unblocks voiceover for screen recordings

That's ~50 minutes of work that takes you from "lots of risk" to "ready to ship."

---

What I'm tracking that you may not have given me yet

• Calendar page rebrand: I still don't know which calendar page you mean (GHL widget? doctorhutcheson.com calendar? Cal.com embed?). Drop me the URL and I'll audit.
• S3 bucket scope: still unclear if it's for patient files (PHI, needs HIPAA setup), brand assets, or customer media. Different answer per use case.
• The Heap → scan integration: you mentioned "fires once Heap response." Need to know: Heap event triggers what action? Where does the data land?
• DocuSign / signature stamp: you mentioned wanting electronic signatures or a saved-signature image stamp on receipts. Want me to wire this in next?

Drop me answers on any of these and I'll ship them.