| T1 | Slack app doesn't exist yet — needs one human with a Slack login (5 min: create from /slack/manifest.yml, install, copy Client ID) | No live Slack play; no corporate pilot surface; /slack install link can't be generated until then | Alexei (or anyone) does the 5-min setup → paste Client ID at /slack → permanent Add-to-Slack link exists | 🟥 |
| T2 | WhatsApp Cloud API group messaging — likely 1:1-only [VERIFY] | Production architecture: group-native bot may be impossible on official API; per-player DM redesign needed at pilot | Verify with Meta at pilot application; design DM-fallback now (docs/14 notes the migration shape) | 🟥 |
| T3 | Bots only run while superlocal is on — no always-on host. (27 Aug: minirig install attempt blocked — box offline ~96d, Tailscale node key expired; needs power-on + tailscale up --force-reauth first, see docs/16) | Demo reliability for Ben/Nico; "is the bot up?" becomes the story | Short-term: move bot processes to minirig (always-on, systemd user services — pattern exists). Long-term: small VPS or worker | 🟥 |
| T4 | No auth / identity — names are free-typed, per-device state | Multi-device play, real accounts, anti-cheat identity, payments | apps/api exists; add magic-link auth (Resend is live) at MVP | 🟧 |
| T5 | App ↔ bot state split — localStorage vs .data JSON, no sync | A crew created in the app can't be played by the bots yet (code binding is manual) | Wire app → apps/api ← bots (the API is built and tested; migration is mechanical) | 🟧 |
| T6 | AI key balance — primary Z.AI key exhausted; running on secondary | AI features (guide, narrator, taunts) die if secondary runs dry | Monitor usage; top up Z.AI; fallbacks already degrade gracefully | 🟧 |
| T7 | CF token lacks Tunnel:Edit — Pages works, named tunnel doesn't | Can't expose the LOCAL server (live bots, live /api/state) at rwf.qalarc.com | Widen token scope or cloudflared tunnel login once; then bots' live data can flow through the same domain | 🟧 |
| T8 | ~~GitHub Actions runners not assigning~~ RESOLVED — transient GitHub incident, not billing (26–27 Aug): runs for e03bc09/ffba9fe queued forever (runner_id: null, no logs), but the very next push (5091488, 26 Aug 16:12) got a runner and deployed green; a fresh workflow_dispatch on 27 Aug also got a runner and succeeded in <60s. Stuck e03bc09 run is a harmless zombie (cancel API returns 409; ignore or delete via UI). | None — push-to-deploy works again; site verified live (200, /system fresh) | Nothing required. Watch item: repo is private so Actions minutes are paid — if it recurs, check github.com/settings/billing → raise spending limit, make repo public (Actions free), or self-host a runner on minirig | 🟩 |