phase: 12_nut_self_test_verdict

Final state verified. Report:

**Verified (phase 12 — all 3 tasks already complete; this was the verification pass)**
- Re-ran the wire capture live: `ups.test.result` = `No test initiated` (apc) / `Done and passed` (cyberpower) — matches `LIVE_SELF_TEST_WORDS` + both fixtures; `/api/summary` has **no** self-test row, `/api/ups/snapshot` reads `cyberpower ok | Done and passed`, `apc unknown | …run a battery test`; both screenshots read back correct.
- Fixed a defect found in prior work while checking the "unrelated live noise" note: `NutMonitor.collect()` reused one upsd socket after a *partially-read* reply, so a 4 s stall produced bogus `expected 'BEGIN LIST VAR', got 'RW apc battery.charge.low "10"'` errors (same hazard on `INSTCMD`, where a late `OK` could be read as a different command's success). `NutClient` now marks a socket whose reply stopped early as `UNUSABLE`, naming the command that stalled; a complete `ERR …` line still leaves it usable (the per-UPS tolerance `collect()` depends on). +3 hermetic pins against the fake upsd; live re-check (09:33–09:44) now logs one line per stall, no cascade.

**Tests / lint / coverage (all green)**
- `bash .agents/validate.sh` → **exit 0**; `uv run pytest --cov --cov-fail-under=94` → **1172 passed, 11 skipped**, coverage **98.63 %** (floor 94); `uv run ruff check src tests agent` clean; `uv run pyright` 0 errors; `node --check` clean. Modified modules: `sources/nut.py` 95 %, `collectors.py` 97 % (no new line uncovered).
- `uv run pytest -m live --live` → **8 passed / 3 failed**, all three the pre-existing `driveripper` agent-transport failures AGENTS.md records (verified still true: unit `inactive`, no installed file, nothing on `:9120`).
- `./scripts/dev-server.sh status`: `nut ok`, `llm:* ok`, `host/disks:deskwork|gateway ok`; the two `driveripper` jobs DOWN.

**Completion criteria**
- Live capture recorded + asserts `ok` — met (re-confirmed from the wire today).
- `/ups` renders `ok`, overview silent, screenshots present — met.
- pytest / cov / ruff / pyright / validate — met (above).
- No other `health.checks[]` behaviour changed — met (diff confined to the `test_result` branch; all prior nut/ups-contract tests unmodified).

**Documentation**: README verified consistent (self-test wording, config table cross-checked against `config.py`, all documented commands/flags exist) — no changes needed beyond phase tasks' edit. AGENTS.md fixes: stale test count (1140 → 1172), `nvme0n1` power-on hours 19 158 → 19 160, gap 4 store-error claim corrected (last stored `store-error` 2026-10-18 06:33, none in the previous 24 h), new `collectors.py`/`sources/nut.py` socket-hygiene note, new gap-1 bullet on the upsd stall, plus live evidence appended to `data/logs/NOTES-self-test.md`.

**Deviations**: made no live *changes* (read-only probes only) — `driveripper`'s agent install needs root (`ops` has no installer/visudo grant) and a battery test stays the operator's call; the upsd 4 s stall itself is upstream-side. `uv pip list --outdated` still advisory (certifi/fastapi/starlette) — dependencies untouched this phase.

**Next pending phase**: `13_llm_page_and_decode_rate`.

README.md

@@ -534,7 +534,13 @@ auth.py ── verifier + session key, loaded once at startup (no row and no see
   `severity` dict, with the electrical readings (line in, output, battery volts and amps,
   battery °C) kept below them in smaller muted type because they are what you read *while* the
   UPS is on battery. Health checks, the battery-test buttons (with their `DANGER`
-  confirmation) and the NUT variable dump are unchanged, just second.
+  confirmation) and the NUT variable dump are unchanged, just second. The `Last self-test`
+  check grades the **outcome** of the driver's word and not its phrasing: a pass
+  (`Done and passed`, the sentence this fleet's CyberPower answers) reads `ok` and raises no
+  overview warning, a fail reads `fail`, and a pack that has never been tested —
+  `No test initiated` or no variable at all — says the same "run a battery test" sentence as
+  `unknown`. Only a word the map has never seen stays a warning, printed verbatim so it can
+  be added.
   The four event feeds (`/system`, `/disks`, `/llm`, `/ups`) are a **paged window** over the `event`
   table rather than a printout of it: the list is a bounded box that scrolls, 50 events to a
   page, with `1–50 of 200` between `‹ prev` and `next ›` (disabled at the ends) underneath.
@@ -614,8 +620,10 @@ battery glyph (`charge %`, `load %`, `runtime left`, `battery_voltage`, the glyp
 verbatim as `tests/fixtures/nut_<ups>.vars` by `./scripts/capture-fixtures.sh`, with the
 device serials scrubbed — and again against live upsd. The captures earn their place: the
 CyberPower answers `battery.mfr.date: CPS`, the APC's `battery.date` is a factory
-placeholder, and **neither** publishes `battery.temperature` or `battery.charger.status`, so
-a hand-written variable table would have tested nobody's memory. `tests/slots_contract.py`
+placeholder, **neither** publishes `battery.temperature` or `battery.charger.status`, and the
+two disagree on `ups.test.result` (the CyberPower has run a test and answers `Done and passed`,
+the APC still answers the factory `No test initiated`), so a hand-written variable table would
+have tested nobody's memory. `tests/slots_contract.py`
 pins gufo's `/slots` rows the same way — hermetically from
 `tests/fixtures/slots_gufo.json` (the verbatim `/slots` reply, captured whenever
 `GUFO_BASE_URL` is set; the idle-slot rows are the shape the grid must handle) and again