actop 1.9.12

Fixed

• l now says why the layout did not change on a narrow terminal. The
  grid preset auto-degrades to stack below 96 columns of dashboard width
  so two columns never squeeze to ~40 cols each. That guard is correct, but it
  was silent: on an 80-col window (or a 140-col one with the process table open,
  which leaves the panel at ~66) pressing l flipped the requested preset
  internally while the screen never changed and nothing explained it — the key
  read as broken. l now raises a toast reporting the layout actually applied,
  and when a requested grid was degraded it names the requirement and the
  measured width (grid needs >=96 cols (panel has 80)). Matches how t
  already reports the theme. The in-app help (?) documents the width rule too,
  and HardwareDashboard.grid_min_width is exposed so the app can explain the
  degrade without reaching into the widget.

Changed

• PyPI publishing moved out of CI/CD into the local tag cut.
  .github/workflows/publish-pypi.yml is deleted and OIDC Trusted Publishing is
  no longer used; scripts/tag_release.sh now builds the sdist + wheel and
  uploads them with twine immediately after pushing the tag and creating the
  GitHub Release, so cutting and publishing a release are one act on the
  maintainer's machine. Credentials come from TWINE_USERNAME / TWINE_PASSWORD
  in the environment. The script pre-flights the credentials, the .venv, and
  the build/twine modules before it pushes anything, because a
  credential failure discovered after the tag is pushed would leave a tagged,
  released, unpublished version that cannot be re-cut. It also wipes dist/
  before building (the upload glob would otherwise ship stale artifacts),
  verifies the built filenames match the tag, and runs twine check before
  uploading. New release optional-dependency group (build, twine), kept out
  of dev so Linux CI stays lean. Maintainer-facing only — no change to the
  published package.

• record-tui-gif skill: prose re-synced with the bundled tape. SKILL.md
  still described a pre-v1.9 stack-based recipe and contradicted
  actop-demo.tape's own comments: it recommended Width 1400 (which the tape
  documents as silently degrading grid to stack mid-take), a Sleep 6s warm-up
  (the tape covers the full ~330s chart span, noting 7s fills 2% of the chart),
  Height 1050/900 (actual: 930), an off-camera l grid->stack switch and an
  l -> g -> t choreography the tape never performs, plus a stale keybinding
  table (t labelled processes, p labelled pause). An agent following it would
  have set the known-broken width and warm-up, pressed l on camera against a
  grid-sized Height, and used the wrong keys. The prose now matches the tape,
  states explicitly that the hero take is grid start-to-finish and that l must
  not be pressed on camera, and lists the real bindings. The skill description no
  longer claims an on-camera layout toggle. No change to the tape itself.