Closing v0.12.0: Surfacing rates_updated_at in synlynk status

The Broader Goal at the End of the Previous PR

PR #264 closed epic #210 in full — every dollar figure the Vizor Effort & Cost tab renders is now either structurally sourced or visibly flagged as an estimate. That PR's closing note named the one thing still standing between the current state and marking the v0.12.0 Measurement Ledger Hardening theme shipped: issue #259, synlynk status surfacing rates_updated_at. Fable's 2026-07-12 strategic review had asked for the _MODEL_RATE_TABLE move out of code and into an updatable data file "with a rates_updated_at shown in status." The data-file move itself shipped back in Measurement Ledger Phase 1 (.synlynk/model_rates.json, PR #236) — synlynk/costs.py's _load_model_rates() has returned a rates_updated_at key ever since. Nothing downstream ever read it. synlynk status — the command most people actually run — had no way to tell a caller whether the rate table behind every cost number on screen was current or a hardcoded fallback from before any rate file existed.

Strategic Shifts in This PR

None to scope. Two design decisions were resolved during brainstorming, both taken as the recommended option without pushback: the new RATES line sits immediately next to the existing BUDGET line rather than getting its own section header, since both lines answer "how trustworthy are the numbers on this screen" and belong together; and a missing/None rates_updated_at renders as an explicit ⚠ warning ("RATES never updated ⚠ (hardcoded defaults)") rather than a silent "unknown" — matching epic #210's own precedent that unmeasured or estimated data must be visibly flagged, never blended in unmarked.

What This PR Shipped

A single-file change to synlynk/status.py, small enough to dispatch as one task rather than the data/rendering split #258 needed:

• New helper — _format_rates_line(rates_updated_at: Optional[str]) -> str returns "RATES updated {date}" when a date is present, else the warning string above.
• _format_status_terminal() signature gained one new keyword parameter, rates_updated_at: Optional[str] = None — a default rather than a required positional arg, specifically so the two pre-existing direct-call tests that don't pass it (test_format_status_terminal_structure, test_format_status_json_valid) kept working unmodified.
• Terminal output inserts the RATES line right after BUDGET in the existing lines list.
• JSON output gains a rates_updated_at top-level key in the payload dict, serializing as the date string or null — json.dumps handles None natively, no special-casing needed.
• cmd_status() now calls synlynk.costs._load_model_rates().get("rates_updated_at") and threads the result through to _format_status_terminal().

Task 1 (dispatched to Codex, job-48d58abb) implemented the change exactly as specified in the plan — verified via git diff against the plan's parent commit to confirm zero deviation, not just a self-report. Five new tests landed in tests/test_ecosystem_status.py: two on _format_status_terminal() directly (date-present and warning-path), one on the JSON payload, and two on cmd_status() end-to-end (rate file present with a date, and no rate file at all asserting null). All 5 passed on independent re-run; the full test_ecosystem_status.py + test_status.py + test_cost_ledger.py set (147 tests) stayed green.

Task 2 (Claude, verification-only, not dispatched) ran the full project suite rather than just the touched files — 1140 passed, 2 skipped, matching the pre-#259 baseline (1135) plus exactly the 5 new tests. This full-suite discipline is a direct lesson from PR #264's CI failure, where a regression in a completely different test file's own fixture (test_vizor_efficiency.py's separate make_test_db()) only surfaced because CI ran everything — a narrowly-scoped local check would have missed it. A manual smoke test confirmed both terminal and JSON output render correctly against a real (non-mocked) _load_model_rates() call on a workspace with no rate file, and a grep audit confirmed _format_status_terminal() has exactly one production call site (cmd_status()), so no other caller needed updating.

Brainstorm Visuals Used

None — this was a small, text-only design resolved through two direct multiple-choice questions (line placement, null-handling), no mockups warranted.

What This Achieved on the Path to Autonomy

This closes the v0.12.0 Measurement Ledger Hardening theme's last remaining item. Combined with #258/PR #264, epic #210 is now fully shipped: cost data is structurally sourced end to end, from the token-extraction adapters (Codex/Claude/Agy/Grok structured-output basis) through the ledger schema (cost_source/estimate_basis provenance columns) through both display surfaces that show cost data to a human — the Vizor web HUD and the terminal synlynk status command. Every number a person looks at now either says where it came from or flags that it's an estimate.

It's also the sixth PR in a row where all production code was dispatched (Codex/Agy/Claude/Grok rotating), with Claude's role held to design, plan-writing, independent diff verification, full-suite test re-execution, and cherry-picking — the PM/reviewer discipline holding even on the smallest scoped task in this run.

Strategic Note: The Goal at the End of This PR

With #258 and #259 both shipped, project-docs/roadmap.md's v0.12.0 Measurement Ledger Hardening arc gets marked shipped next. #260 (Savings Ledger), #261 (dispatch-path unification), #262 (surface consolidation), and #263 (stage constants staleness, filed during #258's side investigation) remain filed and scoped for a later point release, none blocking. #202 (synlynk jobs stale status display) also remains open, tracked separately.