RTW.xyz COCKPIT · INTERNAL
snapshot just now

Cockpit

Initiatives, tasks, web presence, delivery, and financial state — auto-collected from the source of truth, not retyped.

216
actionable
18
blocked
81
done
+0
last 6d
1
hot
Next Action
as of 2026-05-10

Deploy Decision Log API

~30 min

source: NEXT_STEPS.md / Step 1 · target: DigitalOcean droplet (co-locate with Vouch_Man) at decision-log.rtwadvisors.com

Single highest-leverage action. Decision Log API v0.3.0 is code-ready and OAuth works locally. Once on a real host, the agent can publish, the QBO production cutover becomes possible, and §3-§7 of GO_LIVE_CHECKLIST become actionable.

unblocks 26 downstream tasks
  • →GO_LIVE §3 Decision Log integration (5 tasks)
  • →GO_LIVE §5 First production run (5 tasks)
  • →GO_LIVE §7 Zapier orchestration (4 tasks)
  • →QBO production cutover (HANDOFF_QBO_API_FIX.md, 1 task)
  • →First Bookkeeper run on RTW books (NEXT_STEPS.md Step 3, 11 tasks)
Then

  • 2.Deploy Vouch_Man (solis-bot)

    ~1 hr

    11 tasks unblocked

  • 3.Drop credentials.json into solis/ + run Phase 1

    ~30 min

    3 tasks unblocked

Actionable Now· 216
no blocker; do these whenever you've got a free 10 min
  1. Bookkeeper Agent

    GO_LIVE_CHECKLIST.md↗

    § 1. Environment

    4
    • 1..env populated with AWS_REGION, AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY (Bedrock-routed inference per RTW HIPAA BAA policy — superseded direct Anthropic API on 2026-05-10)
    • 2.Bedrock model access enabled in target AWS account/region for BEDROCK_MODEL_ID (default us.anthropic.claude-sonnet-4-5-20250929-v1:0) — request via Bedrock console → Model access if not already enabled
    • 3..env populated with real TAXDOME_API_KEY (when integration is ready)
    • 4.ZAPIER_TELEGRAM_WEBHOOK configured on the Decision Log API side (Vouch_Man bot wired up)
  2. Bookkeeper Agent

    GO_LIVE_CHECKLIST.md↗

    § 3. Decision Log integration (deployed)

    2
    • 5.Decision Log API v0.4.0+ deployed (read-only QBO Reports endpoints). Smoke: curl /api/health returns "version":"0.4.0"; python scripts/pull_qbo_snapshot.py --month 2025-01 --label smoke succeeds.
    • 6.Audit log entries created for every decision, approval, and write
  3. Bookkeeper Agent

    GO_LIVE_CHECKLIST.md↗

    § 4. Vouch_Man Bot — human approval interface

    8
    • 7./start in Telegram returns the welcome message with chat_id
    • 8./status returns "Decision Log API: online" + "Auth: OK"
    • 9./pending lists pending decisions from the deployed Decision Log API
    • 10./review <id> shows the formatted decision card with [Approve][Reject] buttons
    • 11.Tapping Reject updates approval_status='rejected' in the API
    • 12.journalctl -u solis-bot shows no raw bot tokens (only bot<REDACTED> if anything) *(systemd unit name retained as solis-bot even after Vouch_Man rename)*
    • 13.Bot survives sudo systemctl reboot without manual intervention
    • 14.(Optional, separate config) ZAPIER_TELEGRAM_WEBHOOK set on the Decision Log so auto-approved decisions push proactive notifications via Zapier
  4. Bookkeeper Agent

    GO_LIVE_CHECKLIST.md↗

    § 5. First production run (single client)

    5
    • 15.Pick one real client to start. Recommendation: a dermatology practice (highest-confidence sub-vertical per the spec).
    • 16.Run the agent against their actual previous-day QBO export
    • 17.Review every categorization manually — do not trust auto-approval yet
    • 18.Document any miscategorizations in docs/FAILURE_MODES.md (create the file if missing)
    • 19.If accuracy < 95%, do not roll out further — iterate the prompt and bump version
  5. Bookkeeper Agent

    GO_LIVE_CHECKLIST.md↗

    § 6. First 10 production runs (any clients)

    4
    • 20.Continue manual review of every output for the first 10 production runs across any clients
    • 21.Track miscategorizations and their causes; look for patterns
    • 22.If a pattern emerges (e.g., a sub-vertical edge case), bump the prompt (v1.0 → v1.1) with the fix and re-validate
    • 23.After 10 clean runs, transition to spot-check sampling (10% of outputs per the build standards)
  6. Bookkeeper Agent

    GO_LIVE_CHECKLIST.md↗

    § 7. Zapier orchestration

    4
    • 24.Zapier Workflow #1 (QBO Daily Export → Bookkeeper Agent) configured per 14_Zapier_Workflow_Catalog.md
    • 25.Zap fires daily at 7:00 AM PT and successfully calls the agent
    • 26.Output is published to the Decision Log automatically (--publish flag in the Zap's command)
    • 27.Telegram alerts fire on any flagged transactions
  7. Bookkeeper Agent

    GO_LIVE_CHECKLIST.md↗

    § 8. Monitoring

    3
    • 28.Logs from production runs captured somewhere durable (Decision Log API logs, or separate log aggregation)
    • 29.Alerting on agent failures — a cron run dying silently must page someone
    • 30.Weekly review of confidence-tier distribution: are we over-flagging? under-flagging? Adjust thresholds if a clear pattern emerges (rare; usually the prompt should change instead)
  8. Repo root

    BOOKKEEPER_V1_FINISH_PLAN.md↗

    § Phase 3 — Decision Log integration (only if decision 2 = yes)

    1
    • 31.Confirm notify_via_telegram actually fires (Vouch_Man bot needs its .env per session handoff — that's a blocker on this phase). If Vouch_Man credentials aren't ready, build the integration but stub the webhook URL so v1.
  9. Repo root

    HANDOFF_SOLIS_FIX.md↗

    § External setup required (before the new session starts)

    3
    • 32.Telegram chat ID for Askia's account — the bot's "admin chat ID" so it knows where to send notifications. The session can show how to obtain this with /start once the bot is rebuilt.
    • 33.Droplet access — same DigitalOcean droplet as Decision Log API (co-located).
    • 34.Decision Log API key — the same DECISION_LOG_API_KEY value the Decision Log uses. The session needs this to wire bot → API auth.
  10. Repo root

    HANDOFF_SOLIS_FIX.md↗

    § What "done" looks like

    5
    • 35.Bot deployed as a systemd service on the droplet, Active: running
    • 36.Bot survives reboot (sudo systemctl reboot test)
    • 37./pending in Telegram successfully lists pending decisions from the deployed Decision Log API
    • 38.Clicking Approve in /review updates approval_status to 'approved' in the API
    • 39.journalctl -u solis-bot shows clean logs with no bot token visible (regex-redacted)
  11. Repo root

    NEXT_STEPS.md↗

    § Gather before you start (5 min)

    4
    • 40.Droplet access. SSH into your DigitalOcean droplet works: ssh root@YOUR_DROPLET_IP. Use the Vouch_Man droplet — the Decision Log co-locates cleanly. If Vouch_Man isn't deployed yet, stand up a fresh $5/mo Ubuntu 22.04 dr
    • 41.Subdomain decision. Pick one. Recommendation: decision-log.rtwadvisors.com. Create a DNS A record at your registrar pointing it at the droplet IP. Wait for propagation (usually < 5 min, max 1 hour).
    • 42.API key. On your local machine: openssl rand -hex 32. Save the output in your password manager labelled "Decision Log API key". You'll paste it into two places later (the droplet env file and agent-bookkeeper/.env). The
    • 43.Confirm Caddy is installed on the droplet. which caddy should return a path. If not, the DEPLOY.md prereqs section has the install command.
  12. Repo root

    NEXT_STEPS.md↗

    § Wire the agent (5 min)

    3
    • 44.Open agent-bookkeeper/.env
    • 45.Set DECISION_LOG_API_URL=https://decision-log.rtwadvisors.com
    • 46.Set DECISION_LOG_API_KEY=<the openssl key from above>
  13. Repo root

    NEXT_STEPS.md↗

    § Tick the box

    1
    • 47.Edit agent-bookkeeper/GO_LIVE_CHECKLIST.md step 3, mark items checked, commit.
  14. Repo root

    NEXT_STEPS.md↗

    § Step 2 — Vouch_Man dependency (~ blocked, separate workstream)

    6
    • 48.SSH to droplet
    • 49.Edit /etc/decision-log-api/decision-log-api.env
    • 50.Add ZAPIER_TELEGRAM_WEBHOOK=<the Zapier webhook URL>
    • 51.sudo systemctl restart decision-log-api
    • 52.Test: send a decision via the smoke script; confirm Telegram notification arrives
    • 53.Tick GO_LIVE_CHECKLIST step 4
  15. Repo root

    NEXT_STEPS.md↗

    § Pre-flight (20 min)

    2
    • 54.Audit RTW's QBO chart of accounts. Live validation showed the test fixture COA was missing Travel, Professional Dues, CME — the agent correctly refused to force-fit transactions into accounts that didn't exist. Before ru
    • 55.Pull a previous-day or previous-week transaction sample. Since the QBO API integration isn't working yet, this is manual:
  16. Repo root

    NEXT_STEPS.md↗

    § Manual review (30-45 min)

    5
    • 56.Is the category right for RTW's COA?
    • 57.Is the QBO account number right?
    • 58.Is the confidence sensible?
    • 59.If flagged (< 0.85), is the reasoning legitimate?
    • 60.Did the agent miss any context that's obvious to you as the firm's accountant? (those are v1.1 prompt-tuning opportunities)
  17. Repo root

    NEXT_STEPS.md↗

    § Publish to Decision Log (5 min, only after review passes)

    4
    • 61.List pending decisions:
    • 62.For each, confirm the categorization is right, then approve via API:
    • 63.Reject the bad ones similarly via /reject endpoint
    • 64.Tick GO_LIVE_CHECKLIST step 5
  18. project-trident

    EXECUTION-CHECKLIST.md↗

    § Week 1: June 3-9, 2026

    19
    • 65.Create "Tax Preparation — .ai Subscribers" pipeline in TaxDome
    • 66.Create 6 tax prep tags:
    • 67.Create intake organizer (Stage 1)
    • 68.Create 6 tax prep proposals (Stage 2)
    • 69.Test workflow: submit test intake → verify correct proposal fires
    • 70.Complete GBP updates using gbp-copy-paste-checklist.md:
    • 71.Add 5 services with pricing
    • 72.Add appointment booking URL (Calendly)
    • 73.Add rtwadvisors.ai URL
    • 74.Add secondary categories (Tax Consultant, Bookkeeping Service)
    • 75.Create first post (healthcare bookkeeping announcement)
    • 76.Upload photos (at minimum: logo + Askia headshot)
    • 77.Copy review request link
    • 78.Go to YouTube Studio: https://studio.youtube.com/video/a9IaT72MgRU/edit
    • 79.Create thumbnail (or use AI-generated from metadata prompt)
    • 80.Upload thumbnail
    • 81.Schedule publish: June 4, 2026 @ 2:00 PM ET
    • 82.Change visibility from "Unlisted" to "Public" (or keep scheduled)
    • 83.Verify video is in "Healthcare CPA Answers" playlist
  19. project-trident

    EXECUTION-CHECKLIST.md↗

    § (Replace placeholder with real TaxDome organizer link)

    8
    • 84.Update src/data/site.ts with actual tax prep organizer URL
    • 85.Build and deploy to production
    • 86.Test: Click "Request Tax Prep" CTA on /pricing → lands on TaxDome organizer
    • 87.Test: Complete organizer as fake client → verify proposal fires
    • 88.Check YouTube analytics for Week 2 video (first 48 hours):
    • 89.Check Google Business Profile insights:
    • 90.Check TaxDome pipeline:
    • 91.Adjust: Fix any broken links, confusing organizer questions, or proposal routing issues
  20. project-trident

    EXECUTION-CHECKLIST.md↗

    § Week 2: June 10-16, 2026

    15
    • 92.Generate scripts for Week 4 + Week 5 videos
    • 93.Review compliance flags
    • 94.Render in Synthesia (or queue for batch rendering)
    • 95.Upload to YouTube (unlisted)
    • 96.Create thumbnails
    • 97.YouTube Studio: https://studio.youtube.com/video/NW6IKIpb85E/edit
    • 98.Upload thumbnail
    • 99.Schedule publish: June 11, 2026 @ 2:00 PM ET
    • 100.Set visibility to Public
    • 101.Send Welcome Email #1 (immediate after EL signature)
    • 102.Request QBO accountant access
    • 103.Send document request organizer
    • 104.Respond to onboarding questions within [SLA based on tier]
    • 105.Review conversion funnel: Where are prospects dropping off?
    • 106.Adjust: Improve weak link (better CTA, clearer organizer, faster proposal)
  21. project-trident

    EXECUTION-CHECKLIST.md↗

    § Week 3-4: June 17-30, 2026

    10
    • 107.Every Monday: Produce 2 new videos (maintain 4-week buffer)
    • 108.Every Wednesday @ 2 PM: Publish 1 video
    • 109.Every Friday: GBP post announcing latest video + CTA
    • 110.Daily: Check TaxDome for new organizer submissions
    • 111.Within 4 hours: Send proposal to new inquiries
    • 112.Day 2: Follow up if no proposal response
    • 113.Day 5: Final follow-up ("Is this still something you're interested in?")
    • 114.By 5th of month: Deliver all monthly closes (existing clients)
    • 115.Within 7 days of signup: Complete new client onboarding
    • 116.Within 24-72 hours: Respond to client messages (based on tier SLA)
  22. project-trident

    EXECUTION-CHECKLIST.md↗

    § July 1: Yvette Handoff Begins (Week 5)

    4
    • 117.Week 1 (July 1-7): Yvette shadows Askia on 3-5 consultation calls
    • 118.Week 2 (July 8-14): Yvette leads calls with Askia on standby
    • 119.Week 3 (July 15-21): Yvette owns calls, Askia handles escalations only
    • 120.Week 4 (July 22-28): Yvette full ownership, weekly debrief with Askia
  23. project-trident

    EXECUTION-CHECKLIST.md↗

    § August 4: Mid-Sprint Check-In

    6
    • 121.Current MRR: $_____ / $80K goal (___%)
    • 122..ai subscribers: ___ (target: 35-40)
    • 123.Churn rate: ___% (target: <5% in first 90 days)
    • 124.Video metrics:
    • 125.Review requests sent: ___
    • 126.Google reviews: ___ (target: 10+ by sprint end)
  24. project-trident

    EXECUTION-CHECKLIST.md↗

    § August 11: Final Push Week

    4
    • 127.Review all open proposals in TaxDome
    • 128.Follow up with prospects who went cold
    • 129.Deliver all August monthly closes on time (build trust for renewals)
    • 130.Prep September content batch (maintain momentum)
  25. project-trident

    EXECUTION-CHECKLIST.md↗

    § Monday: Content Production

    3
    • 131.Produce 2 new videos (agent-content pipeline)
    • 132.Review prior week's video metrics
    • 133.Adjust topics if certain videos underperform
  26. project-trident

    EXECUTION-CHECKLIST.md↗

    § Tuesday: Client Delivery

    3
    • 134.Respond to all client messages from weekend
    • 135.Advance any jobs stuck in pipeline stages
    • 136.Prep upcoming monthly closes (if approaching 5th of month)
  27. project-trident

    EXECUTION-CHECKLIST.md↗

    § Wednesday: Video Publish Day

    3
    • 137.2:00 PM ET: New video goes live
    • 138.Post to GBP announcing video
    • 139.Share on LinkedIn / X if applicable
  28. project-trident

    EXECUTION-CHECKLIST.md↗

    § Thursday: Sales Follow-Up

    4
    • 140.Check new TaxDome inquiries (organizer submissions)
    • 141.Send proposals to new leads
    • 142.Follow up with warm prospects (no response to proposal)
    • 143.Update CRM or tracking sheet
  29. project-trident

    EXECUTION-CHECKLIST.md↗

    § Friday: Metrics & Planning

    4
    • 144.Review week's numbers (signups, churn, video views, GBP engagement)
    • 145.Identify bottlenecks (where are prospects dropping off?)
    • 146.Plan next week's priorities
    • 147.30-min debrief with Yvette (once she's onboarded)
  30. project-trident

    EXECUTION-CHECKLIST.md↗

    § Must-Have (Non-Negotiable)

    4
    • 148.$60K+ MRR (75% of $80K goal — shows strong trajectory even if not fully hit)
    • 149.25+ new .ai subscribers (demonstrates product-market fit)
    • 150.<10% churn rate (proves delivery quality and value perception)
    • 151.13 videos published (Week 2-14 of content calendar, proving content machine works)
  31. project-trident

    EXECUTION-CHECKLIST.md↗

    § Nice-to-Have (Stretch Goals)

    5
    • 152.$80K MRR (100% of goal)
    • 153.40+ new .ai subscribers
    • 154.<5% churn rate
    • 155.10+ Google reviews (4.5+ star average)
    • 156.Top 3 local pack for "Healthcare CPA Atlanta"
  32. project-trident

    EXECUTION-CHECKLIST.md↗

    § Red Flags (Indicates Need for Pivot)

    4
    • 157.<$40K MRR by Week 9 (Aug 4) → Major course correction needed
    • 158.>15% churn rate → Delivery or expectation mismatch issue
    • 159.<20% conversion rate → Pricing, positioning, or sales process broken
    • 160.Askia working >70 hours/week consistently → Not sustainable, need to hire faster
  33. project-trident

    gbp-copy-paste-checklist.md↗

    § Completion Checklist

    8
    • 161.Business description updated (healthcare focus)
    • 162.5 services added with pricing
    • 163.Appointment booking URL added
    • 164.rtwadvisors.ai URL added
    • 165.Categories updated (Tax Consultant + Bookkeeping Service)
    • 166.First post published
    • 167.Photos uploaded (at least logo if available)
    • 168.Review link copied and saved
  34. Marketing collateral

    PLAYBOOK_HANDOFF.md↗

    § 11. Success criteria

    13
    • 169.agent-content/CANONICAL_VIDEO_SETTINGS.md — locked config with canonical avatar UUID
    • 170.rtw-marketing/PLAYBOOK.md — full methodology document
    • 171.rtw-marketing/variants/marketing.md — fully designed (mirror Episode 10)
    • 172.rtw-marketing/variants/tax_review.md — fully designed worked example
    • 173.rtw-marketing/variants/monthly_recap.md — frontmatter stub
    • 174.rtw-marketing/variants/quarterly_estimate.md — frontmatter stub
    • 175.rtw-marketing/variants/ad_hoc.md — frontmatter stub
    • 176.rtw-marketing/variants/year_end_recap.md — frontmatter stub
    • 177.rtw-marketing/variants/onboarding.md — frontmatter stub
    • 178..claude/skills/rtw-video/SKILL.md — executable skill
    • 179.Clients/Real Results/.../client_review_video/scenes.yaml — rewritten to tax_review variant
    • 180.agent-content/DRIFT_DETECTED.md — only if avatar UUID drift confirmed; else omit
    • 181.Final report message to Askia summarizing findings + open questions
  35. Repo root

    rtw-qbo-quarterly-audit-checklist.md↗

    § Pre-audit setup

    3
    • 182.Identify all active QBO instances under RTW management
    • 183.Open this checklist; date-stamp the audit
    • 184.Create a quarter-specific findings file at rtw-policies/qbo-quarterly-audits/YYYY-QQ.md
  36. Repo root

    rtw-qbo-quarterly-audit-checklist.md↗

    § 1. Customer list review

    5
    • 185.Open Customer list for each QBO instance
    • 186.Scan all active customers for patterns suggesting individuals rather than payers/entities (first-last name combinations)
    • 187.Flag any patient entries
    • 188.Remove flagged entries (merge into appropriate payer/entity entry if needed)
    • 189.Document any findings in the audit file
  37. Repo root

    rtw-qbo-quarterly-audit-checklist.md↗

    § 2. Memo field scan (last 90 days)

    4
    • 190.Run "Transaction Detail" or "Transaction List" report for last 90 days
    • 191.Search memo and description columns for PHI signals:
    • 192.Flag and redact each occurrence
    • 193.Document count + types found
  38. Repo root

    rtw-qbo-quarterly-audit-checklist.md↗

    § 3. Attachment audit

    5
    • 194.Browse Documents tab or run attachment report
    • 195.Open each attachment uploaded in the last 90 days
    • 196.Confirm no EOBs, Superbills, statements with patient detail, or other PHI-containing files
    • 197.Remove and replace with PHI-free versions if found
    • 198.Document count + types found
  39. Repo root

    rtw-qbo-quarterly-audit-checklist.md↗

    § 4. Bank feed review

    4
    • 199.Sample 20 transactions from last 90 days (random selection across all bank/credit accounts)
    • 200.Confirm transaction descriptions are payer/aggregate-only
    • 201.Confirm no patient-identifying detail in saved transaction notes
    • 202.Document any slips
  40. Repo root

    rtw-qbo-quarterly-audit-checklist.md↗

    § 5. Vendor list review

    2
    • 203.Confirm vendor list is practice-vendor only (no patients)
    • 204.Flag and remove any patient entries
  41. Repo root

    rtw-qbo-quarterly-audit-checklist.md↗

    § 6. Findings & remediation log

    4
    • 205.Summarize total findings by category (memo / attachment / customer / vendor / bank-feed)
    • 206.Document remediation steps taken
    • 207.Identify root cause if slips happened (training gap? bank-feed import bug? new staff member?)
    • 208.Save findings file under rtw-policies/qbo-quarterly-audits/YYYY-QQ.md
  42. Repo root

    rtw-qbo-quarterly-audit-checklist.md↗

    § 7. Policy reinforcement

    3
    • 209.If findings indicate a slip-up pattern, schedule a re-training session for anyone touching QBO
    • 210.Update rtw-qbo-phi-exclusion-policy.md if a control gap was identified
    • 211.Note in rtw-hipaa-risk-assessment.md if residual risk needs to be re-evaluated
  43. Repo root

    rtw-qbo-quarterly-audit-checklist.md↗

    § 8. Sign-off

    5
    • 212.Audit completed by: _____________________
    • 213.Date: _____________________
    • 214.Total findings: _____________________
    • 215.Remediated: Y / N
    • 216.Follow-up actions: _____________________

Tasks

216 actionable · 18 blocked · 81 done
blockers: 18 Solis creds · 30m
  • project-trident ▸ 96 actionable

    EXECUTION-CHECKLIST.md ↗

    16/112
    6d ago
    • 5/5
      Content & Marketing
    • 3/3
      Tax Prep Pipeline
    • 3/3
      Onboarding & Handoff
    • 4/4
      Infrastructure
    • 1/20
      Week 1: June 3-9, 2026 ▸ now
    • 0/8
      (Replace placeholder with real TaxDome organizer link) ▸ now
    • 0/15
      Week 2: June 10-16, 2026 ▸ now
    • 0/10
      Week 3-4: June 17-30, 2026 ▸ now
    • + 11 more sections
  • Repo root ▸ 35 actionable

    rtw-qbo-quarterly-audit-checklist.md ↗

    0/35
    24d ago
    • 0/3
      Pre-audit setup ▸ now
    • 0/5
      1. Customer list review ▸ now
    • 0/4
      2. Memo field scan (last 90 days) ▸ now
    • 0/5
      3. Attachment audit ▸ now
    • 0/4
      4. Bank feed review ▸ now
    • 0/2
      5. Vendor list review ▸ now
    • 0/4
      6. Findings & remediation log ▸ now
    • 0/3
      7. Policy reinforcement ▸ now
    • + 1 more sections
  • Bookkeeper Agent ▸ 30 actionable

    GO_LIVE_CHECKLIST.md ↗

    15/45
    23d ago
    • 3/7
      1. Environment ▸ now
    • 6/6
      2. Live Claude validation (4 fixtures) — PASSED 2026-05-04
    • 4/6
      3. Decision Log integration (deployed) ▸ now
    • 2/10
      4. Vouch_Man Bot — human approval interface ▸ now
    • 0/5
      5. First production run (single client) ▸ now
    • 0/4
      6. First 10 production runs (any clients) ▸ now
    • 0/4
      7. Zapier orchestration ▸ now
    • 0/3
      8. Monitoring ▸ now
  • Repo root ▸ 25 actionable

    NEXT_STEPS.md ↗

    0/25
    29d ago
    • 0/4
      Gather before you start (5 min) ▸ now
    • 0/3
      Wire the agent (5 min) ▸ now
    • 0/1
      Tick the box ▸ now
    • 0/6
      Step 2 — Vouch_Man dependency (~ blocked, separate workstream) ▸ now
    • 0/2
      Pre-flight (20 min) ▸ now
    • 0/5
      Manual review (30-45 min) ▸ now
    • 0/4
      Publish to Decision Log (5 min, only after review passes) ▸ now
  • Marketing collateral ▸ 13 actionable

    PLAYBOOK_HANDOFF.md ↗

    0/13
    15d ago
    • agent-content/CANONICAL_VIDEO_SETTINGS.md — locked config with canonical avatar UUID
    • rtw-marketing/PLAYBOOK.md — full methodology document
    • rtw-marketing/variants/marketing.md — fully designed (mirror Episode 10)
    • + 9 more open
  • Repo root ▸ 8 actionable

    HANDOFF_SOLIS_FIX.md ↗

    7/15
    24d ago
    • 2/5
      External setup required (before the new session starts) ▸ now
    • 5/10
      What "done" looks like ▸ now
  • project-trident ▸ 8 actionable

    gbp-copy-paste-checklist.md ↗

    0/8
    6d ago
    • Business description updated (healthcare focus)
    • 5 services added with pricing
    • Appointment booking URL added
    • + 4 more open
  • Repo root ▸ 1 actionable

    BOOKKEEPER_V1_FINISH_PLAN.md ↗

    30/31
    25d ago
    • 6/6
      Phase 0 — Repo scaffold (blocking; nothing runs without this)
    • 7/7
      Phase 1 — Make existing tests pass + close low-effort gaps
    • 8/8
      Phase 2 — Standards conformance
    • 4/5
      Phase 3 — Decision Log integration (only if decision 2 = yes) ▸ now
    • 5/5
      Phase 4 — Pre-launch QC + bump to v1.0
  • Repo root

    HANDOFF_SOLIS_PERSONA_BUILD.md ↗

    1/19
    29d ago
    • 1/8
      External setup required (before the new session starts) ⛔ Solis creds
    • 0/11
      What "done" looks like ⛔ Solis creds
  • Repo root complete

    HANDOFF_QBO_API_FIX.md ↗

    12/12
    23d ago
    • 5/5
      External setup required (before the new session can write code)
    • 7/7
      What "done" looks like
Blockers come from src/data/blockers.json section map. Override per task with inline [BLOCKED: key] tag. File names link to VSCode.

Build Projects

11 tracked

  • Bookkeeper Agent

    agent-bookkeeper bedrock ✓

    v1.1 tagged 5/4 — first run on RTW books pending

    stalled 16d ago

  • Tax Preparer Agent

    agent-tax-preparer bedrock ✓

    ~30% scaffold; gated on Bookkeeper live run

    stalled 24d ago

  • Decision Log API

    decision-log-api

    v0.3.0 — not deployed

    stalled 16d ago

  • Inbox Brain

    inbox-brain gemini

    Phase 0→1: ingest in progress

    stalled 24d ago

  • Vouch_Man (Decision Log bot)

    solis-bot

    v0.2.0 code-ready, not deployed

    stalled 24d ago

  • Solis (persona bot)

    solis

    Phase 0 cleanup done; Phase 1 blocked on credentials.json

    missing missing

  • rtwadvisors.com

    rtwadvisors-com

    Astro/Vercel — Elite 20 funnel

    stalled 27d ago

  • rtwadvisors.ai

    rtwadvisors-ai

    Astro/Vercel — scaled tier

    < 1w 6d ago

  • Content automation

    agent-content bedrock ✓

    CLI: topic.md → script + AEO metadata

    today today

  • Marketing collateral

    rtw-marketing

    Strategy + 52-wk content calendar

    stalled 15d ago

  • Phone System (Quo + Sona AI)

    phone-system

    .ai 470 live; .com 404 porting from Grasshopper

    stalled 24d ago

Web Presence

2 sites
AEO checks live: /llms.txt, /sitemap.xml, JSON-LD — probed each npm run sync

Client Delivery

manual

  • Hatchery Planning

    Mon/Tue/Fri in-person, 9:30a–1:30p ET

    next: 05-11 3 upcoming

  • Synergy HomeCare (Hiram)

    Weekly Tuesday 2p ET

    next: 05-12 3 upcoming

  • Synergy HomeCare (Villa Rica)

    Monthly first Tuesday 12:30p ET

    next: 06-02 1 upcoming

  • Real Results (Medical Weight Loss)

    Monthly CFO advisory

    next: 06-02 1 upcoming

  • Travel windows

    build cadence drops to zero

    • 2026-05-15 → 2026-05-17
      Chicago
      Bruno hold
    • 2026-05-22 → 2026-05-25
      South Haven, MI
      Memorial Day
    • 2026-05-29 → 2026-05-31
      Nashville, TN
      Hee Haw
Source: MS Graph credentials not set (MS_GRAPH_CLIENT_ID + MS_GRAPH_TENANT_ID); using manual snapshot

Financial KPIs

stubbed
  • MRR
    —
  • AR aging
    —
  • Elite 20 capacity used
    —
  • Pipeline value
    —
Blocker

QBO env not set (QBO_CLIENT_ID, QBO_CLIENT_SECRET, QBO_REALM_ID, plus a token DB) OR decision-log-api not yet deployed with completed initial OAuth

Memory State

44 entries

  • feedback-monthly-reporting-no-deep-research

    unknown

    Monthly reporting package production must not use the deep-research skill — too slow and too expensive for a repeatable monthly workflow

    2d old

  • reference-deepseek-mcp

    unknown

    "deepseek-mcp server — DeepSeek-R1 on AWS Bedrock with PHI pattern gate, registered user-scope; PHI-forbidden by default"

    6d old

  • project-rtw-ai-site-trident-alignment

    unknown

    "rtwadvisors.ai reconciled to Project Trident 2026-05-26 — 3 verticals, new tiers/pricing, all-tiers self-serve via TaxDome (no Stripe); two decisions intentionally amend the locked Trident report"

    8d old

  • project-rtw-brain-consolidation

    unknown

    The RTW Brain — single model-neutral source of truth across all LLM CLIs; consolidation plan + locked decisions

    8d old

  • rtw-marketing (Strategy + Compliance Docs)

    project

    Marketing strategy hub — compliance cheat sheet (GA Rule 20-12-.15), pricing baseline, 52-week content calendar, brand template. Plain markdown, no code. Lives at RTW_Claude_Projects/rtw-marketing/.

    8d old

  • project-taxdome-mcp-build

    unknown

    taxdome-mcp (Playwright GUI automation + TaxDome Drive doc I/O) — Phase 0+1+2+2.5 done (10 tools incl screenshot/wait/table_read), needs Bedrock-lane live test

    8d old

  • project-telegram-foreman-bot

    unknown

    "Deferred build: Telegram bot as foreman/supervisor for long-running Claude automation sessions (taxdome-mcp, bookkeeper agent, etc.) — progress pings, blocker escalation, approval gates"

    8d old

  • reference-client-report-video-skill

    unknown

    client-report-video skill — branded client-delivery videos walking a client through their reports; thin layer over the agent-content Synthesia pipeline; confidentiality-gated

    8d old

  • google-drive-distribution

    unknown

    Google Workspace Business Plus w/ HIPAA BAA signed 2026-05-26; Drive is compliant output distribution layer accessible from any device

    8d old

  • reference-outlook-mcp

    unknown

    "Outlook MCP server — Microsoft Graph calendar + mail (search/move/delete/mark/batch), registered user-scope in Claude Code"

    8d old

  • reference_voice_io_claude_code

    unknown

    "Local voice input (WhisperWriter STT) + spoken output (Stop-hook TTS) for Claude Code on Askia's Surface Laptop 6"

    8d old

  • reference-working-capital-forecast-skill

    unknown

    working-capital-forecast skill — RTW-branded cash-reserve/distribution-capacity forecast PDFs; client-facing sibling of tax-workpaper-build

    8d old

  • feedback-home-games-block-travel

    unknown

    "When planning travel/cruises, treat Atlanta Falcons home games AND Georgia Tech home football games as hard blockers, not just calendar noise"

    9d old

  • rtw-pdf-branding-standard

    unknown

    "All RTW client/firm deliverables (workpapers, letters, reports) must render as PDF with RTW branding and logo — markdown is source, PDF is the deliverable."

    10d old

  • openclaw-cockpit

    unknown

    "OpenClaw 2026.4.15 configured as a multi-provider chat playground (Claude, Gemini, OpenAI/Codex, DeepSeek, Ollama) — explicitly scoped non-PHI. Locked 2026-05-21."

    13d old

  • provider-sandbox-env-cleanup

    unknown

    Open security cleanup — RTW_GEMINI_PROJECTS/.env and RTW_VIBE_Projects/.env contain live QBO + TaxDome production tokens alongside non-BAA LLM provider keys. Discovered 2026-05-21.

    13d old

  • project-trident-sprint

    unknown

    "90-day sprint to $80K MRR (target 2026-08-18) via .ai-led acquisition across 3 productized verticals (Healthcare / Professional Services / Trades & Contractors), supported by new tax preparer hire and existing offshore bookkeeper Rona"

    13d old

  • reference-image-gen-stack

    unknown

    "Image generation stack across RTW projects — Bedrock-hosted Stability and Amazon Titan/Nova image models, region routing (us-west-2 default for text-to-image), cost notes, BAA caveat"

    13d old

  • reference-ocr-stack

    unknown

    "OCR/image-processing stack across RTW projects — engine routing by doc type and HIPAA path, install paths, and code snippet templates"

    13d old

  • rtw-provider-sandbox-folders

    unknown

    "Six parallel `RTW_<Model>_Projects` folders under `C:\\Users\\AskiaRoberts\\` — one per LLM provider — for context-switching between provider CLIs."

    13d old

  • palisades-violations-posture

    unknown

    "HOA Palisades Violations Committee operates as an off-ramp, not an enforcement arm — CMA does the letters; the committee proactively reviews to find amicable resolutions before hearings/fines escalate"

    15d old

  • hoa-palisades

    unknown

    "Askia's HOA roles + governing-document landmarks for Homeowners' Association of Palisades, Inc. (396 units, Paulding County GA, managed by CMA)"

    15d old

  • project-client-oviedo-household

    unknown

    Oviedo Household client — multi-entity book (6 entities) with folder rollup at Clients/Oviedo Household; Entity Map is the source of truth before any Kim call

    16d old

  • rtw-ent-llc-entity-profile

    unknown

    "Legal-entity facts for RTW ENT LLC (the firm) — confirmed from the filed 2024 1120-S. EIN, S-corp election, sole shareholder, cash basis, NAICS 541219, vehicle inventory, and the $398K shareholder suspended-loss carryforward into 2025."

    16d old

  • Decision Log API capabilities at v0.4.6 — what writes back, what doesn't

    reference

    Decision Log API v0.4.6 supports QBO write-back for Purchase / Bill / JournalEntry / Deposit. Invoice / SalesReceipt do NOT — their line accounts come from ItemRef, not AccountRef. Multi-line / split transactions also unsupported. Read endpoints include GeneralLedger (added v0.4.6).

    16d old

  • reference-motion-mcp

    unknown

    "Motion MCP server — direct API integration built 2026-05-13, registered in Claude Code user scope"

    21d old

  • Bookkeeper Agent — Jan 2025 production catch-up status

    project

    First production publish of the Bookkeeper Agent (RTW's own books, Jan 2025). 152 of 229 transactions written to QBO; remainder pending/blocked/rejected. Pipeline proven for Purchase/Bill/JournalEntry/Deposit.

    22d old

  • AWS Bedrock keys live in agent-content/.env (shared across all RTW Claude agents)

    reference

    AWS_ACCESS_KEY_ID + AWS_SECRET_ACCESS_KEY for Bedrock are in agent-content/.env; reuse them when wiring up new Bedrock-based agents. AWS_REGION there is us-east-2 but model access works in us-east-1 too.

    22d old

  • PHA Agent Build Project Context

    project

    Year 1 agent workforce architecture, dual firm model (PHA + Elite 20), tech stack, and build philosophy

    23d old

  • All Claude API through AWS Bedrock

    feedback

    Architectural rule — every Claude API call from RTW-owned code routes through AWS Bedrock, not direct Anthropic API, for HIPAA BAA chain coverage

    24d old

  • agent-content (RTW Content Pipeline)

    project

    Lightweight CLI script — topic.md → Claude API script → compliance regex sweep → AEO YouTube metadata. Human-in-loop, NOT autonomous. Feeds Synthesia render via Zapier MCP. Lives at RTW_Claude_Projects/agent-content/.

    24d old

  • Inbox Brain Project

    project

    Email-first AI assistant — Gemini Flash classify-at-ingest + Gemini Pro RAG queries over a SQLite ledger. Fully deployed 2026-05-10 to DigitalOcean droplet. Lives at RTW_Claude_Projects/inbox-brain/ locally; production at /opt/inbox-brain/ on 157.230.218.51.

    24d old

  • RTW.xyz Internal Cockpit

    project

    Personal status dashboard at rtwadvisors.xyz — tracks build projects, web presence, client delivery, financials. Astro/Vercel, brand-aligned with .com/.ai. Lives at RTW_Claude_Projects/rtwadvisors-xyz/.

    24d old

  • HIPAA Compliance Posture

    project

    RTW's HIPAA architecture — QBO PHI Exclusion Policy (Path A+C) + risk assessment + quarterly audit cadence; locked 2026-05-10

    24d old

  • RTW Phone System Build

    project

    Two-line Quo (formerly OpenPhone) Business setup with Sona AI on both numbers, Grasshopper port in progress, Smith.AI deferred to day-60 pending real call volume data

    24d old

  • RTW Video Production Model — Locked 2026-05-10

    project

    Locked broadcast-explainer production model for all RTW video content — multi-scene Synthesia with custom branded slide backgrounds, eighth-grade comprehension, TTS pronunciation rules

    24d old

  • Two-Bot Split (Vouch_Man vs Inbox Brain bot)

    project

    Disambiguates the two Telegram bots — Vouch_Man (approval bot, code-ready, NOT deployed) vs Inbox Brain (live on droplet via @RTW_Advisors_MoltBot). The legacy solis/ persona form-bot is dead.

    24d old

  • RTW Services Droplet

    reference

    DigitalOcean droplet hosting RTW production services. First real droplet for the practice — created 2026-05-10. Runs inbox-brain + decision-log-api + Vouch_Man + Caddy.

    24d old

  • RTW Web Presence Project Context

    project

    Two-site marketing strategy (rtwadvisors.com = Elite 20 funnel, rtwadvisors.ai = scaled templated/avatar tier), AEO-led, locked 2026-05-05

    25d old

  • RTW Video Stack — Locked 2026-05-08

    project

    Locked video tooling decision for RTW marketing engine — Synthesia primary spokesperson, Higgsfield deferred for B-roll, Zapier MCP for orchestration

    26d old

  • Pay-in-advance billing model for .ai (Shelbie feedback)

    feedback

    Family-member real-user feedback on the .ai pricing model and tone — pay in advance, underpromise/overdeliver on response times.

    27d old

  • CPA License — Georgia

    user

    Askia is licensed as a CPA in Georgia. Advertising/solicitation governed by GA State Board of Accountancy Rule 20-12-.15; client confidentiality by Rule 20-12-.11.

    27d old

  • Webflow Designer was unintuitive — bailed to Gamma

    feedback

    User abandoned Webflow for Gamma because the Designer's visual GUI didn't match his mental model. Bias toward code-first or template-driven options when recommending CMS platforms.

    29d old

  • Email Stack — Microsoft 365

    user

    Askia's primary inbox aroberts@rtwadvisors.com is on Microsoft 365 / Outlook (NOT Gmail). The rtwadvisors.com tenant is on Azure / Entra ID. Use Microsoft Graph + MSAL for any email integration.

    29d old