`meetings/2026-05-13_090002-morning.md`

57 lines
# DW Morning Stand-Up — Wed 2026-05-13 09:00 PDT

---

## 1. Carryover from Last Meeting

⏳ **Steve / Doctor `apply-templates.js` decision** — Day 15. Lawyer pipeline still fully gated. No movement.
⏳ **Steve / `dw_admin` PG rotation** — Eight consecutive stand-ups. Runbook at `~/Projects/_shared/docs/pg-rotation-runbook.md` is ready. 32-minute task.
⏳ **Crash cluster** — `glassbeadedwallpaper`, `wallco-ai`, `thesetdecorator`, `whatsmystyle`. Counters reset overnight (restarts now show 98 / 12 / 14 / absent) but no root-cause diagnosis was done. They are crashing again.
⏳ **Freddy port rotation** — Six days overdue since soak ended 2026-05-07. Runbook ready.
⏳ **Gemini keys** — Both still hardcoded. `Letsbegin` wired 7 Gemini routes 20 min ago — but the keys themselves haven't been minted/rotated via `/secrets` yet.

---

## 2. Overnight (Last 12 Hours)

**Crashes / instability:**
- `glassbeadedwallpaper` — 98 restarts (counter reset, still crashing). Day 18 undiagnosed. Highest-risk process in fleet.
- `wallco-ai` — 12 restarts. Counter reset but active.
- `thesetdecorator` — 14 restarts. Counter reset.
- `whatsmystyle` — absent from crash list this morning; unclear if fixed or just restarted and not yet re-accumulated.

**Long-running jobs:** Fleet of 40+ core agents stable at 11h uptime, 0 restarts — Norma social, visual/AI factory, pd-crawler, pd-dedupe, animals-ingest all healthy.

**Git activity:**
- 10 DW era-sites (`1800s`–`1980swallpaper`) — duplicate `<header>` removal shipped ~2 hours ago. ✅
- `Letsbegin` — cost-tracker helper + 7 Gemini route hooks committed 20 min ago. ✅
- Multiple snapshot commits (Bertha, Designer-Wallcoverings, HERMES, Ken, Norma, NationalPaperHangers) — routine automated snapshots.
- No unexpected remote commits from collaborators.

---

## 3. Today's Plan + Agent Assignments

- `debugger-agent / glassbeadedwallpaper` — pull last 200 pm2 log lines, identify crash signature — Day 18, highest restart count in fleet — named root cause (OOM / missing dep / port conflict)
- `debugger-agent / wallco-ai` — same log pull + root cause — counter reset means this will re-accumulate fast — named root cause or stop recommendation
- `debugger-agent / thesetdecorator + whatsmystyle` — triage both in same pass — both have been crashing for days without diagnosis — stop-or-patch decision for each
- `secrets-manager` — route Gemini key #1 (`AIzaSyAO0…QejMo`) via `/secrets` after Steve mints it — `Letsbegin` already wired 7 routes, key still hardcoded — key stored in registry, hardcoded value removed
- **Steve** — mint Gemini key #1 + #2 at aistudio.google.com and paste via `/secrets` — unblocks Letsbegin + Patty routes — done in < 5 min
- **Steve** — Doctor `apply-templates.js` decision — Day 15, lawyer pipeline fully gated — yes/no/modify
- `freddy-agent` — run `rotate-pg-scoped.sh` for Freddy port — soak ended 6 days ago, runbook ready — port rotation complete

---

## 4. Blockers & Decisions Steve Owes

- **Doctor `apply-templates.js`** — yes/no/modify. 15 days without lawyer pipeline progress.
- **`dw_admin` PG rotation** — 32-minute runbook, Steve must initiate. Eight stand-ups overdue.
- **Gemini key mint** — two keys, five minutes at aistudio.google.com, unblocks multiple agents.
- **Norma Lane B vs Lane A** — DB migration deferred; needs explicit per-action auth before `rotate-pg-scoped.sh` can run.

---

## 5. One Thing to Think About

The crash cluster has now survived two full days of planned triage that never happened — and the counters reset overnight means the data is getting harder to read, not easier. Is there a specific reason these four agents haven't been stopped while diagnosis runs? Stopping a crashing agent costs nothing; leaving it running 100+ times a day may be masking downstream damage.