docs: add a project status snapshot

[twistedmelonman]

Adds docs/STATUS.md, plus a pointer to it from CLAUDE.md.

Why

The backlog design says what each item is and why. Nothing said where the project actually stands — so resuming has meant re-deriving it from issues, commits, and live API queries each time. This session did exactly that twice.

What it records

• Layer status. Foundation partly done (F1, F4), identity done (I3), fleet's W0 done, L and N not started.
• The critical path is now W1 → W2 → W3. I3 and W0 have left it, and W2 inherits a largely conformant fleet rather than one it must protect from scratch — so it needs re-scoping against current state, not the design's original numbers. The design's W2 pilot choice (scripts, picked because its check was "Pro-gated and therefore not enforced") expired at I3 and needs re-picking.
• What to pick up next, in priority order.
• What is deferred by explicit decision. This is the part that matters most: Part D on TILSIT/MIMOLETTE, the shared-token split, dev-env staying public, and the parked photo-game-poc token copy are all decisions, not gaps. Without recording that, the next session re-opens each one.
• The two disciplines this project keeps paying for — resolve the thing rather than matching its label, and validate every fix against a known-bad case — with this session's instances as live examples.

Verification

Every factual claim was measured on 2026-09-05 rather than recalled, and the doc says so with a standing instruction to re-measure before acting:

• alias present at ~/.config/bash/gh-wrapper.sh:188
• docs/runbooks/org-migration-rename.md exists
• org-migration suite: 38 tests pass
• issue states: [infra] Migrate smartwatermelon/ repos to a new GitHub org for org-level secret management #54 closed; org-migration: three conditions the design's failure table does not cover #84, [infra] Make smartwatermelon/scripts public: secret audit, history rewrite, branch protection #85, [infra] Four repos have no enforced review check (needs a review workflow first) #89, [infra] Evaluate org rulesets as the scaling mechanism for review enforcement #90 open
• fleet counts re-audited after the W0 writes, not taken from the write responses

Markdown only. Codebase and adversarial reviewers clean.

https://claude.ai/code/session_01MM2jf6c7eNN4QMq32GdC5X