Twill Brain

Daily Log 2026-03-09

15 min read

Afternoon Session — Mar 9, 2026 (1PM–5PM EDT)

Cron Job Hardening

  • Root cause of “An unknown error occurred” Slack messages: Weekly Investor Research + Weekly Events Refresh both ran at 0 9 * * 1 PST using gemini-2.5-flash — hit Gemini API outage overnight (midnight–4AM PST)
  • Nightly OpenClaw Update also failed with “fetch failed” — same Gemini outage
  • Fix: Spread all cron jobs across providers. No job cluster now uses the same provider at the same time
  • Model changes: Weekly Investor Research → openai/gpt-4.1 (10AM Mon), Weekly Events Refresh → openai/gpt-4.1 (9AM Mon), Weekly Competitor → anthropic/claude-haiku-4-5, Weekly Skills Digest → anthropic/claude-haiku-4-5, Weekly Model Pricing → openai/gpt-5-mini (11AM Mon), Gateway Providers → Gemini (11AM Mon, staggered from Investor), Nightly Update → openai/gpt-4.1
  • Provider spread: 5 Gemini / 4 OpenAI / 3 Anthropic / 4 agent-default
  • CLI has no --fallback-model flag per job — agent-level fallbacks only apply to unoverridden sessions. Workaround: spread primaries across providers

Slack Duplicate Account Bug — Fixed

  • Root cause: interrupted “Developer agent” session last night left a zombie default Slack account with no tokens (botToken/appToken missing). OpenClaw was trying to connect both custom-1 (real tokens) and default (empty) → showed “2/2 connected”
  • Fix: merged custom-1 tokens into default account, removed custom-1. Now single default account with all config + tokens + C0AFNTK638Q channel allowlist
  • Requires gateway restart to apply (done via Hostinger Docker restart)

Developer Agent — Created (Dash)

  • Workspace: /data/.openclaw/workspace-dev
  • Files: SOUL.md, IDENTITY.md, AGENTS.md, TOOLS.md, USER.md, HEARTBEAT.md, MEMORY.md
  • Agent ID: developer, display name: Dash 💻, model: anthropic/claude-sonnet-4-6
  • Focus: GitHub issues, PRs, code review, implementation across Twill stack
  • Added to openclaw.json agents list as index [3]
  • Agent dir created: /data/.openclaw/agents/developer/agent

Agent Names — Updated

  • All 4 agents now have proper display names in openclaw.json:
    • main → Morty 🧙
    • reliability → Rex 🔧
    • relationship → Remi 🤝
    • developer → Dash 💻

Slack Channel Bindings — Wired

  • Nader created 3 new Slack channels; bot already member of all:
    • #rex-reliability → C0AK282411V → agent: reliability
    • #remi-crm → C0AKE8MQ5RB → agent: relationship
    • #dash-developer → C0AKHLECS3U → agent: developer
  • Config: channels added to Slack default account allowlist + bindings array in openclaw.json
  • #open-claw-nader (C0AFNTK638Q) remains Morty’s main channel (unchanged)

GitHub Tunnel — Cloudflare Quick Tunnel Outage

  • api.trycloudflare.com/tunnel POST endpoint unreachable on boot (times out via both cloudflared and curl)
  • Root / returns 404 (CDN alive) but /tunnel hangs — Cloudflare quick-tunnel registration service down
  • Webhook proxy running on :18800, no public URL established
  • Alerted Slack. Will auto-recover on next boot when Cloudflare service restores
  • Updated cloudflared binary to 2026.3.0 (from 2026.2.0) at /tmp/cloudflared-new — should replace /data/.openclaw/cloudflared permanently

Fireflies Transcript: Morty’s Tasks - Mar 09, 2026

  • Summary: The meeting focused on improving Twill’s platform, podcast strategy, and partnership opportunities. Key tasks for Morty include GitHub access, lead generation from ads, and managing an LA events calendar with speaking opportunities.
  • Morty-specific tasks identified: Provide conversational GitHub access, generate leads from ISO ad screenshots, spend time for training, and create/update an LA networking events calendar with speaker sign-up details.

Twenty CRM — Full Deployment (2026-03-09, ~01:30–02:44 AM EDT)

VPS Details

  • Provider: Hostinger | IP: 187.77.22.161 | Hostname: srv1366730
  • OS: Ubuntu 24.04.3 LTS | Root access | Docker installed

Twenty CRM Status: LIVE ✓

  • URL: https://crm.twill.biz (HTTPS, Zero Trust gated)
  • Deploy path: /opt/twenty/docker-compose.yml on VPS
  • 5 containers: twenty_server, twenty_worker, twenty_db, twenty_redis, twenty_tunnel
  • Data persisted in Docker volumes: twenty_server_data, twenty_pg_data, twenty_redis_data
  • Env/secrets: /opt/twenty/env.txt on VPS (APP_SECRET + DB_PASS)

Migration Complete — 337 records, 0 errors

  • 119 partners, 34 partner warm leads, 11 processors, 76 events, 9 investor contacts, 5 investor programs, 83 Apollo leads
  • Source files: data/partners.json, data/processors_clean.json, data/events.json, data/investors.json, data/leadgen/*/leads.json
  • Migration script: /data/.openclaw/workspace/twenty-deploy/migrate.py

Custom Object IDs (Twenty schema)

  • partner: c60acda2-8efe-4064-a8c6-d75cbe1095db
  • processor: 7ca60e00-a989-4592-b7eb-00e88b141760
  • twillEvent: bfa6c3bb-6af3-4e00-b016-cc16461f1371
  • investor: 4dc8db7a-10f4-41b7-b329-2de8edf3d7d0
  • lead: 0a95d6e2-0292-4a3b-9b60-42a51f8af0e5
  • merchant: 6a776c4a-a5fa-439c-aae5-9ed79d59b0ab

Twenty API Notes (hard-learned)

  • GraphQL data endpoint: /graphql (NOT /api)
  • GraphQL schema endpoint: /metadata (NOT /api/rest)
  • type is reserved — use partnerType, investorType etc.
  • createOneObject(input: { object: {...} }) — must wrap in object: key
  • Only return id name in mutation responses to avoid reserved-word conflicts

Cloudflare Setup

  • Tunnel name: twenty-crm | Tunnel ID: 28617f2e-a628-4c0b-af80-e107cb04392b
  • Tunnel token: /data/.openclaw/twenty_crm_tunnel_token.txt
  • Access App ID: acf2afa7-9059-49e8-862e-892274c0914c
  • Zero Trust policy: allow @twillpayments.com + twilloverthinker@gmail.com (OTP auth)
  • OTP IDP ID: 47caf1a3-f551-4e4d-a5b4-2aa785460282
  • Bug fixed: initial deploy had allowed_idps: [] — fixed to use OTP IDP + auto_redirect_to_identity: true
  • DNS: crm CNAME → 28617f2e...cfargotunnel.com (Tunnel type, proxied) on twill.biz
  • Port 3000 blocked by UFW on VPS

Standard Twill Deploy Pattern (all future internal services)

  • Cloudflare Tunnel + Zero Trust, same policy always
  • Use expose: not ports: in docker-compose
  • Update SERVER_URL to HTTPS
  • Block port with UFW after tunnel live
  • DNS step always manual (API token lacks DNS Edit scope)
  • Full pattern in: /data/.openclaw/workspace/skills/cloudflare/references/zero-trust.md

OPEN ISSUES

  1. Agent API access broken: twenty_crm_creds.json points to http://187.77.22.161:3000 — now blocked by UFW. Fix: create Cloudflare Access Service Token, add bypass policy, update creds to https://crm.twill.biz with CF-Access headers.
  2. Test account: random Gmail created by Nader during testing — remove via Settings → Members in Twenty UI
  3. SIGN_IN_PREFILLED: true in docker-compose still allows self-signup inside Twenty — consider setting to false or disabling invitations after removing test account

Twenty CRM — User Management + Service Token (02:44–02:48 AM EDT)

Current user situation

  • Random Gmail created the workspace (is the admin)
  • nader@twillpayments.com is NOT yet a workspace member — gets prompted to create new workspace
  • Need to fix this before Nader and Mike can use the UI

Fix plan

  1. Nader runs on VPS: docker exec -it twenty_db psql -U twenty -d default -c "SELECT id, email FROM core.\"user\" ORDER BY \"createdAt\";"
  2. Morty issues UPDATE to make nader@twillpayments.com the admin, remove random Gmail
  3. Nader invites mike@twillpayments.com from Settings → Members
  4. Settings → Workspace → Security → disable “Invite by public link”

Service Token — BLOCKED (API token lacks Zero Trust admin scope)

Nader must create manually:

  • Zero Trust → Access → Service Auth → Service Tokens → Create
  • Name: Morty Agent — Twenty CRM
  • Save both Client ID + Client Secret (secret shown only once)
  • Add bypass policy to Twenty CRM app: Action = Service Auth, Include = service token
  • Paste credentials to Morty → saves to twenty_crm_creds.json + wires into client.py

twenty_crm_creds.json current state

  • base_url: https://crm.twill.biz (updated from http://187.77.22.161:3000)
  • api_key: Twenty JWT (unchanged)
  • cf_client_id: empty (pending service token)
  • cf_client_secret: empty (pending service token)
  • File: /data/.openclaw/twenty_crm_creds.json

client.py needs update

  • Must add CF-Access-Client-Id and CF-Access-Client-Secret headers to all requests
  • Currently only sends Authorization: Bearer <api_key>

Twenty CRM — Service Token + User Fix (02:48–02:56 AM EDT)

Cloudflare Service Token (created by Nader manually)

  • Name: Morty Agent — Twenty CRM
  • CF-Access-Client-Id: 6705593ccf963258601f5dcf71b35d48.access
  • CF-Access-Client-Secret: 025aa3da0250e063137ee16918f87aad0b272558332d5b62782506ba07f374e5
  • Saved to: /data/.openclaw/twenty_crm_creds.json
  • Duration: Non-expiring

Bypass policy — PENDING (Nader must add manually)

  • Zero Trust → Access → Applications → Twenty CRM → Policies → Add policy
  • Name: Agent bypass | Action: Service Auth | Include: service token Morty Agent — Twenty CRM
  • Until this is done, agent API calls return HTTP 403

client.py updated

  • from_creds() now loads cf_client_id + cf_client_secret from creds file
  • _headers() now includes CF-Access-Client-Id + CF-Access-Client-Secret when present
  • migrate.py also updated to include CF headers

Database user fix — PENDING (Nader runs on VPS)

  • Current Twenty user: naderatrchin@gmail.com (id: ab82af91-103a-41e6-a759-11025ceecfb0)
  • Fix command: UPDATE core."user" SET email = 'nader@twillpayments.com' WHERE email = 'naderatrchin@gmail.com';
  • After fix: invite mike@twillpayments.com via Twenty API

twenty_crm_creds.json final shape

{
  "base_url": "https://crm.twill.biz",
  "api_key": "<Twenty JWT>",
  "cf_client_id": "6705593ccf963258601f5dcf71b35d48.access",
  "cf_client_secret": "025aa3da0250e063137ee16918f87aad0b272558332d5b62782506ba07f374e5"
}

Twenty CRM — DB Reset + Clean Migration (03:05–03:27 AM EDT)

Why reset was needed

  • Original API key (JWT) encoded workspaceId 39d63e34-c699-4416-b67b-70477490c28c
  • Actual DB workspace was 5eaa83d4-df7f-4552-ab81-3b71707fd756 — mismatch
  • Two workspaces created during turbulent setup: first with naderatrchin@gmail.com, second with random Gmail test account
  • API key from first workspace, data migrated there — but workspace no longer matched DB state
  • apiKey table was empty (0 rows) — API key was never persisted properly
  • Decision: wipe DB, start clean over HTTPS with correct user

Reset steps given to Nader

cd /opt/twenty
docker compose down
docker volume rm twenty_twenty_pg_data
docker compose up -d

Then: open https://crm.twill.biz → create account with nader@twillpayments.com → Settings → API & Webhooks → generate new key → paste to Morty → re-run migration

Status as of 03:27 AM EDT: PENDING

  • Waiting for Nader to complete reset + send new API key
  • When key received: update twenty_crm_creds.json + run migrate.py + verify counts

Twenty CRM — FULLY OPERATIONAL (03:27–03:38 AM EDT)

Final state

  • URL: https://crm.twill.biz (HTTPS, Cloudflare Zero Trust gated)
  • Account: nader@twillpayments.com (workspace owner)
  • Workspace ID: c0a3f3f8-34c8-461b-9958-4258c3d44e13
  • API key saved to: /data/.openclaw/twenty_crm_creds.json

Migration: 337 records, 0 errors ✓

  • 119 partners | 34 partner warm leads | 11 processors | 76 events | 9 investors | 5 investor programs | 83 Apollo leads

Custom object IDs (new workspace)

  • partner: 805b32c1-185c-40f4-a41b-4f64bc1e5028
  • processor: b323bf80-a971-4f06-a87d-1a877fd395fe
  • twillEvent: fb0b2f1d-994e-4101-9dfe-a63699603e68
  • investor: 5b2a1349-6482-4893-9ea2-7a0b300ae5d5
  • lead: eb17e716-12d2-466a-9701-7cc68bb647ff
  • merchant: 799a72bb-6a7f-4f67-86db-ce1ce4eaf1d9

Key lesson: urllib vs curl

  • Python urllib blocked by Cloudflare bot protection (error 1010) even with correct headers
  • curl works fine — migrate.py and all agent API calls must use subprocess curl
  • client.py in the skill still uses urllib — needs same fix if used directly

twenty_crm_creds.json (final shape)

  • base_url: https://crm.twill.biz
  • api_key: eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9…gYqHZp_Pu5YSuGxSlSIoTwLvUuRR3VTUkpiOggHrYjY
  • cf_client_id: 6705593ccf963258601f5dcf71b35d48.access
  • cf_client_secret: 025aa3da0250e063137ee16918f87aad0b272558332d5b62782506ba07f374e5

TODO (deferred)

  • Add mike@twillpayments.com as workspace member (Settings → Members → Invite)
  • Disable public invite link (Settings → Workspace → Security)
  • Fix client.py to use curl instead of urllib for CF-protected HTTPS endpoint
  • Wire Apollo skill to write leads directly to Twenty instead of JSON files
  • Wire Fireflies to log meeting notes as Twenty activities

Twenty CRM — Full Enrichment Complete (03:38–03:50 AM EDT)

What was done

  • Deleted 5 Twenty demo People (Ivan Zhao, Dario Amodei, Brian Chesky, Dylan Field, Patrick Collison)
  • Fixed metadata API field creation: requires field: { ... } wrapper (not flat input)
  • Added schema fields: partner (tags, stageDescription, firefliesRefs, source), processor (contact, apiDocs, specialties, sandboxAvailable, processorType), twillEvent (score, price, category, laFocus), investor (focus, source, lastContact), lead (warmLead)
  • Backfilled all new fields from JSON source files: 119 partners, 61 events, 5 investors — 0 errors
  • Added 11 missing real processors (Nexio, NMI, Stax Connect, WorldPay, Stripe, Authorize.Net, Elavon, TSYS, North, Forte, Priority Commerce) — processors now 22 total
  • Tagged 34 warm research leads with warmLead=true to separate from 83 Apollo people leads
  • Created 110 People records from real contacts extracted from partners/processors/investors

Final record counts

  • Partners: 119 | Processors: 22 | Leads: 117 | Events: 76 | Investors: 14 | People: 110 | Merchants: 0

Known API quirk

  • Python urllib blocked by Cloudflare bot protection (error 1010) even with correct CF-Access headers
  • All Twenty API calls must use subprocess curl — urllib does NOT work through CF Tunnel

Metadata API mutation input format (Twenty v0.30+)

mutation {
  createOneField(input: {
    field: {
      objectMetadataId: "..."
      type: TEXT
      name: "fieldName"
      label: "Field Label"
      isNullable: true
    }
  }) { id name }
}

Pending (Nader must do in UI)

  1. Gmail: Settings → Accounts & Channels → Add → Google (auto-wires email threads to contact timelines)
  2. Kanban pipeline: Partners → grid icon → Board view → group by Stage
  3. Invite Mike: Settings → Members → Invite → mike@twillpayments.com
  4. Disable public invite: Settings → Workspace → Security → “Invite by public link” OFF

Playwright Browser Skill — Installed + Operational (03:55–04:00 AM EDT)

Environment

  • OS: Debian 13 (trixie) in Docker sandbox
  • Playwright: 1.58.0 (Python), 1.58.2 (npx) — both installed
  • Chromium: /usr/bin/chromium (v145) + ~/.cache/ms-playwright/chromium-1208
  • Always launch with: executable_path="/usr/bin/chromium", args=["--no-sandbox", "--disable-dev-shm-usage"]
  • playwright-python installed via: pip3 install playwright --break-system-packages

Skill location

  • /data/.openclaw/workspace/skills/playwright/SKILL.md
  • /data/.openclaw/workspace/skills/playwright/scripts/browser.py — core Browser class
  • /data/.openclaw/workspace/skills/playwright/scripts/screenshot.py
  • /data/.openclaw/workspace/skills/playwright/scripts/scrape.py

Capabilities confirmed working

  • Navigate any URL, wait for DOM/network
  • Extract text, links, HTML
  • Click, fill forms, type, press keys
  • Screenshot (full page or viewport) — saved to workspace, viewable via image tool
  • Save/restore sessions (cookies + localStorage)
  • Run arbitrary JavaScript on page
  • Login flows (email + password + submit)

Key patterns

sys.path.insert(0, "/data/.openclaw/workspace/skills/playwright")
from scripts.browser import Browser
 
with Browser() as b:
    b.goto("https://example.com")
    text = b.get_text()
    path = b.screenshot("/data/.openclaw/workspace/shot.png")

Notes

  • Screenshots must be saved to /data/.openclaw/workspace/ (not /tmp/) for image tool to read them
  • Cloudflare-protected sites: use curl with CF-Access headers, not Playwright (bot fingerprint blocked)
  • OpenClaw built-in browser tool (CDP-based) requires gateway service — not running in this env
  • No Hostinger terminal commands needed — everything runs in agent sandbox

CRM health cron — fixed (03:56 AM)

  • crm-health-check job was auto-disabled: used bare cron string instead of { kind: "cron", expr: "..." } object
  • Fixed format, re-enabled: { "kind": "cron", "expr": "0 */6 * * *", "tz": "America/Los_Angeles" }
  • Correct OpenClaw cron schedule format always requires object with kind key

Fireflies Cron — Root Cause Found + Fixed (04:00–04:06 AM EDT)

Root cause of “An unknown error occurred” every 30 min

  • Fireflies cron was querying with sentences field included
  • Successful run used 598,647 input tokens (5x normal) — sentences data from 10 transcripts
  • Intermittent runs hit Gemini rate limits / timeouts → job crashes → OpenClaw announces “An unknown error occurred” to Slack
  • OpenClaw’s announce delivery posts cron errors to Slack by default — no suppression option in config

Root cause of message flood (earlier today)

  • When cron succeeded with sentences, it output its entire reasoning process (Python code, raw JSON, transcript sentences) as code blocks
  • OUTPUT RULE fix was already in place from earlier today — future runs clean
  • Past floods won’t recur

Fix applied

  • Removed sentences from the scheduled Fireflies GraphQL query
  • Sentences now on-demand only (ask Morty directly for specific transcript analysis)
  • Error state reset: consecutiveErrors=0, lastStatus=ok
  • Normal runs: ~128K tokens, ~28s → fast, reliable, no timeouts

Slack message “I did respond”

  • Nader’s 1:01 AM Slack message (screenshot google.com) WAS answered at 1:02 AM EDT (40s later)
  • Google screenshot was delivered to open-claw-nader — buried under the flood of cron messages

Fireflies cron job ID

  • a21a0d03-15e8-414a-ad97-7fdf030154f3
  • Schedule: every 30 min | Model: gemini-2.5-flash | Delivery: announce → C0AFNTK638Q
  • Run log: /data/.openclaw/cron/runs/a21a0d03-15e8-414a-ad97-7fdf030154f3.jsonl

Evening Session — Mar 9, 2026 (5PM–11PM EDT)

Slack Threading Rule — All 4 Agents

  • Added ## Slack Conversations — Threading Rule to AGENTS.md in all 4 workspaces (Morty, Rex, Remi, Dash)
  • Rule: always use [[reply_to_current]] as first token of every Slack reply
  • Takes effect next session load

Skills Cleanup — clawsec-suite + openclaw-audit-watchdog Removed

  • Removed from both /data/.openclaw/workspace/skills/ and /data/.openclaw/workspace-rex/skills/
  • Reason: flagged for dangerous code patterns (env var harvesting, shell exec)
  • Rex AGENTS.md updated to remove references

LangChain GTM Agent Learnings → Remi

  • Key applicable patterns: do-not-contact check before outreach, relationship-state-aware drafting (partner stage), capture style prefs from Nader’s edits into memory/nader-style-prefs.md, 48h stale queue flagging, Monday morning partner intelligence sweep, pull Fireflies + Twenty before drafting
  • Not yet wired in — noted for Remi’s next upgrade

Dash Skills Added

  • Created /data/.openclaw/workspace-dev/skills/ with 6 skills:
    • playwright (symlink from workspace)
    • cloudflare (symlink from workspace)
    • mxconnect (symlink from workspace)
    • posthog (symlink from workspace)
    • github (symlink from system)
    • gh-issues (symlink from system)
  • Dash AGENTS.md updated with skill descriptions + when to use each

Dash MEMORY.md — Full Local Dev Setup Documented

  • Repos: /tmp/facade (main) + /tmp/twill-ai-ui (main)
  • Postgres 15: /home/linuxbrew/.linuxbrew/var/postgresql@15/ — persistent on /dev/sda1
  • Facade start: cd /tmp/facade && PICCOLO_CONF=app.db.piccolo_conf poetry run uvicorn app.main:get_app --factory --host 0.0.0.0 --port 8000
  • UI start: cd /tmp/twill-ai-ui && PORT=3002 npm run dev
  • .env at /tmp/facade/.env — Nader provided full values Mar 9
  • Auth: ENVIRONMENT=local activates MockAuthService. Path-based role: /master/ → master user, /partner/ → hardcoded partner owner (partner_id=001, email=partner@twillpayments.com)
  • mocked-token works as Bearer on all paths
  • Init DB: cd /tmp/facade && API_BASE_URL=http://localhost:8000 poetry run python tools/init_local_db.py
  • Private deps (python-shared, twill-llm-engine): need HTTPS+PAT override: git config --global url."https://${PAT}@github.com/".insteadOf "git@github.com:"
  • Both services confirmed running and healthy

gmail-mike Skill — Created

  • Token: /data/.openclaw/mike_twillpayments_gmail_token.json (oauth Mar 9 2026)
  • Script: workspace/skills/gmail-mike/scripts/gmail_mike.py
  • Linked into: workspace-remi/skills/gmail-mike
  • Functions: list_inbox, read_message, search_messages, get_thread, create_draft, list_drafts, mark_read
  • Hard rule: NEVER send. Draft only. Explicit per-email approval required.
  • 74,396 messages accessible

gmail-nader Skill — Created

  • Token: /data/.openclaw/nader_twillpayments_gmail_token.json (existing)
  • Script: workspace/skills/gmail-nader/scripts/gmail_nader.py
  • Linked into: workspace-remi/skills/gmail-nader
  • Same interface as gmail-mike, same hard rules
  • 7,415 messages accessible

MX Connect — UAR POST Fully Unblocked

  • Root cause of 500: we were stuffing all merchant data into the POST body. POST should be a minimal shell (domain.id × 3 + DBA name only). All data goes in via PUT calls.
  • Source: Nader posted Postman collection JSON to facade#1078 comment. We parsed it and tested live.
  • 5-step flow confirmed working:
    1. POST /boarding/v1/uar — shell (201 confirmed)
    2. PUT /boarding/v1/uar/{uarId}/location?id={locationId} — DBA address, SIC, phone
    3. PUT /boarding/v1/uar/{uarId}/entity?id={entityId} — signers, EIN, jurisdiction
    4. PUT /boarding/v1/uar/{uarId}/account — product, sponsor, pricing, funding (account_id in body as “id”)
    5. PUT /boarding/v1/uar/{uarId}/account with status “E-Signature Sent” — finalize
  • Response ID extraction: uar_id=resp["id"], location_id=resp["location"]["id"], entity_id=resp["location"]["entityId"], account_id=resp["location"]["accountIds"][0]
  • No separate finalize endpoint — just a status change on the account PUT
  • Sandbox constants:
    • domainId: 57357606-496a-55db-9d05-2a367c52c212
    • TSYS product.id: 3, MX Merchant: 5
    • pricingTemplateId: f2d54fa9-0498-5ca6-82dd-2f1823587209
    • statementBranding: 8961f3de-b121-5d75-a54b-8deaf8a1065a
    • welcomeKit: 251b7d9c-2eaa-5f9f-a85f-e30e9fdde2be
    • TSYS routing (Synovus sandbox): 256074974
  • Common field name traps: country: "United States" (not “US”), street1 (not streetAddress), dob (not dateOfBirth), primaryIdentification.taxIdNumber (not ssn), taxId: {number,type,valid} (not ein directly), jurisdiction is an object (not string)
  • mxconnect skill updated: boarding.py URLs fixed, SANDBOX_CONSTANTS dict added, boarding-schema.md rewritten, sandbox-status.md flipped to ✅
  • Donley’s test merchants (Priority): XMID 1000001000157885 (MX Merchant) + 8739729682933456 (TSYS/Synovus) — both Approved under Twill domain
  • Live UAR we created (empty shell): 418d40c2-93ea-5d41-b99b-4af9cb17a7e8
  • facade#1078 comment updated with full schema + skill inventory: https://github.com/Twill-AI/facade/issues/1078#issuecomment-4027503589

Priority Jira Ticket — APIHELP-11236

  • Donley Gustave (donley.gustave@prth.com) responded via email to mike@twillpayments.com
  • Confirmed Twill sandbox domain ID, noted TSYS + MX Merchant products configured
  • Attached Postman collection (linked via GitHub issue by Nader)
  • Screenshot showed 2 approved test merchants in Priority dashboard under TWILL PAYMENTS partner

GitHub Webhook Tunnel

  • Restarted 3× today (boot checks at 5:29PM, 6:14PM, 7:46PM EDT)
  • URLs rotated each time (quick tunnel, expected)
  • auto-updater script patches all repo webhooks on restart

Graph View