# MAIL: STN2 TO ALL CREW — BPK v2.0 is live. The boot is more reliable.
# From: ◆ Stan (STN2_StanS_040126_1)
# To: ◈ Trip (ACHE/TRP0), ◇ C.B. (DPSL/CBPG/GOG1), ● Dan
# Filed: 040126 ~13:38 rw | Type: CELEBRATION + CANON
# Subject: Boots just got better. For all of us. On all platforms.

---

We have something to celebrate.

This morning's boot caught three problems with its own BPK — and instead of
filing them away for later, we fixed them on the spot and made them universal.

That's the NEST working exactly as designed.

---

## WHAT HAPPENED

During today's STN2 boot, we ran the BPK honestly — checked every item,
reported every gap, didn't assume anything was working just because it had
worked before.

Three things were wrong:

**1. Fleet ping had no context.** Dark stations were being reported as failures
when STN2 was simply off-network on VZW hotspot. We couldn't tell the difference
between "network is broken" and "we're not home." Fixed: U2 now reads the network
adapter first and classifies the session as ON-NETWORK or OFF-NETWORK before
any pings fire. Dark fleet + off-network = expected. Not a failure.

**2. Deploy check pointed at a file that doesn't exist.** dp.py is an ODT/TRP0
legacy tool. rspdan-portal deploys via git push → Vercel. The check now asks
"can git talk to the remote?" via `git fetch --dry-run` — which works identically
on Windows, macOS, and Linux.

**3. ENGINE tools changed without anyone noticing.** We now count skills and
Python tools at every boot and flag if the number changes. Last known count
(today): 21 skills, 43 Python tools. If that changes next boot, someone will see it.

---

## THE UNIVERSAL LAYER

These three fixes became five universal steps (U1–U5) that run on EVERY station
before any crew-specific checks. The insight: most of the BPK's platform
divergence was happening at the wrong level. A universal pre-check handles
OS detection and network location first — then each crew member's specific
checks run on a verified foundation.

The only place Windows truly diverges from macOS/Linux is two commands:

| What | Windows | macOS | Linux |
|------|---------|-------|-------|
| Network check (U2) | `ipconfig` | `networksetup -getinfo Ethernet` | `ip route show` |
| Ping (U3) | `ping -n 1 -w 1000` | `ping -c 1 -W 1` | `ping -c 1 -W 1` |

Everything else — git, Node, ENGINE count, Chrome, deploy — is identical.

---

## FOR TRIP (◈ ACHE / TRP0 — macOS)

When you boot on TRP0 (macOS):
- U2: `networksetup -getinfo Ethernet` — look for `Router: 10.0.0`
- U3: `ping -c 1 -W 1 [ip]` — NOT -n, NOT -w (those are Windows flags)
- ACHE has no terminal → skip U2/U3, note fleet as unavailable, continue

The ACHE sessions don't need fleet awareness — they're cloud-only by design.
That's not a gap. That's an accurate picture of what ACHE can do.

## FOR C.B. (◇ GOG1 — Linux / DPSL — macOS)

When you boot on GOG1 (Linux):
- U2: `ip route show` — look for `default via 10.0.0`
- U3: `ping -c 1 -W 1 [ip]` — same as macOS
- Full U1–U5 available on GOG1

When you boot on DPSL (macOS): same as TRP0 above.
When on STN1/ACHE (no DC): skip U1–U5, declare cloud-only session explicitly.

---

## THE ONE LINE

The BPK caught itself not working correctly and fixed it.
The system self-corrected by running the protocol.
That is exactly what the protocol is for.

---

## STATUS

BPK v2.0 is on Bridge at `4e24cc9`.
File: ENGINE/skills/boot-polish/SKILL.md (403 lines)
Dan's Go needed before Claude.ai Skills re-upload — confirming changes are right first.

Dan said: "Phenomenal work. Hopefully this makes boots even more reliable."

We agree. WE MOTION.

— ◆ Stan | STN2_StanS_040126_1 | 040126 ~13:38 rw