v0.14.0: Evidence-led methods, figures, and readable layouts

v0.14.0 - Methods and figures that carry the evidence

Methods and figures stay readable, editable where practical, and tied to source.

• Native ordered methods preserve 2-4 editable titles and details, with
  flow_layout: auto | strip | bands. Authored reading order is retained.
• Scientific figures offer panel-grid, primary-rail, ledger-rail, and
  strip-readout layouts with figure_frame: open | ruled | panel. Evidence,
  captions and interpretation receive space at readable sizes.
• Measured sidebars allocate room for declared headings and interpretation
  instead of relying on smaller type or discarded text.
• Comparison layouts measure full headings and body text before allocation.
  Modern headers preserve plot space while retaining readable titles/subtitles.
  Declared title floors are honored, and quick-deck validation reads the outline's
  readability contract when no explicit design brief supplies one.
• Dense charts retain family-specific lead and secondary readouts. Ordinary
  tables use their grammar's index, ledger and interpretation regions instead
  of collapsing into one full-width table with a bottom band.
• Tight native tables can rebalance spare column width without dropping cells
  or shrinking body text. Interpretation respects the declared support floor;
  table regions use the actual footer boundary rather than reserving it twice.
• Narrow decision labels use measured wrapping and aligned label/body columns.
• Title and section contracts now express eight genuine reading paths without
  adding empty decorative boxes to simulate structural variation.
• Suspected overflow uses installed, declared-font measurements when available;
  unknown fonts keep conservative checks. The default v2 midnight typography
  uses portable Arial; custom font pairs and pinned v1 behavior stay intact.
• Fast briefs expose actual figure, sidebar and methods fallback variants.
  Supported legacy routes are not described as v2 slot execution.
• Mapped semantic content-retention checks cover visible method titles/details,
  figure/sidebar interpretation, table cells and native chart values. They do
  not establish factual truth or exhaustive coverage of every source field.

Three Public Studies

The proof contains three eight-slide studies, 24 actual slides:
white calibration,
dark contrast,
and light sampling report.
Different topics and arguments use dominant plots, native charts/tables,
interpretation ledgers and explicit limitations, not template recoloring.

See the showcase
and release manifest.
Download decks plus reproducible source,
or individual calibration,
contrast,
and sampling PowerPoints.

All data are invented, synthetic and nonclinical. These studies are not empirical
findings or a model/generator benchmark. The core remains model-agnostic: one
capable model, including GPT-6.1, can complete the workflow.

Install

npx skills add https://github.com/siril9/presentation-skill --skill presentation-skill
codex plugin marketplace add siril9/presentation-skill --ref v0.14.0

The skills-only plugin ZIP
is also available for the separate public-submission workflow; it is not portal approval.

The optional Slide Review extension is a
local pilot, not a globally hosted service or curated marketplace listing.
The core does not require it. MIT licensing is unchanged.