@damiangilgonzalez/harness 1.3.7

package/skills/presentations/SKILL.md

@@ -0,0 +1,258 @@
+---
+name: presentations
+description: "Use when building, theming, or exporting a presentation deck — pitch, sales, keynote, board/QBR, leave-behind one-pager — from slide structure to a token-based theme to PDF or editable PPTX (Marp, Slidev, python-pptx). NOT the words (that is `marketing`), NOT the visual tokens (that is `design`), NOT the investor story arc (that is `pitch-deck`)."
+tags: [presentations, pptx, slides, deck]
+recommends: [marketing]
+origin: risco
+---
+
+# Presentations — Stunning, On-Brand Decks (PDF + PPTX)
+
+> Two first-class pipelines, one decision. Design-led Markdown decks (Marp/Slidev) for a deck that
+> *flipa* and exports clean to PDF **and** PPTX, or native editable PowerPoint via `python-pptx`
+> when the user must hand off a `.pptx` people will edit in PowerPoint/Keynote/Google Slides.
+> The deck owns *structure, visual system, and export*. The **words** are `marketing`'s job; the
+> **visual tokens** are `design`'s. This skill orchestrates all three into a finished deck.
+
+Also out of scope: a web-native, scroll/animation-heavy **landing page** (that is
+`../design/SKILL.md` + `../nextjs/SKILL.md`); a **video / motion explainer** as the deliverable — this
+skill uses motion only as restrained slide transitions and builds, never as the medium; and the live
+financial *model* behind an investor deck — this skill renders the slides and cites the model as the
+source of truth.
+
+## Brand grounding (gate — clears before the first headline)
+
+**Never produce deck copy or a deck narrative without a complete brand study.** A deck is the brand on
+stage: generic slides read as "another AI deck" the moment they hit the projector, and the only cure is
+grounding every headline, claim, and tone choice in a real, persisted brand profile. This is the same
+gate `marketing` and `design` enforce — decks share the study, they do not fork it.
+
+1. **Locate the brand study.** Read the project's root `CLAUDE.md` for a `## Brand & voice` section linking into `docs/wiki/brand/` (the `harness` Karpathy-wiki convention: compiled brand articles under `docs/wiki/brand/`, raw inputs the user pastes under `docs/raw/brand/`). No `CLAUDE.md`, no link, or a link that points nowhere = ABSENT.
+2. **Check completeness** against the checklist in `references/brand-grounding.md` — it extends the shared brand checklist with **deck-specific** dimensions: deck purpose, audience & setting, length, presenter-vs-leave-behind, and must-include slides. Any empty dimension = INCOMPLETE.
+3. **If ABSENT or INCOMPLETE, STOP and interview the user** — one focused batch at a time, never all questions at once. Voice samples are mandatory; never fabricate a voice. Then persist: write/update the brand study under `docs/wiki/brand/` (raw inputs verbatim under `docs/raw/brand/`), and add/update the `## Brand & voice` link in root `CLAUDE.md`. Exact format → `references/brand-grounding.md`.
+4. **Only once the study is complete, proceed** — and cite which articles drove the deck (e.g. "narrative grounded in `docs/wiki/brand/value-proposition.md`, voice in `voice.md`").
+
+Single exception: if the user explicitly says "skip it, rough draft", you may produce a clearly-labelled
+`DRAFT (ungrounded — not brand-checked)` and still recommend running the gate before it ships.
+
+## Design the message before the pixels
+
+A deck is a *communication* artifact, so the design that matters most is the message design. Everything
+below descends from **audience-centered design** — every choice serves the audience's understanding, not
+the presenter's comfort.
+
+**Plan the message (six questions).** Lock these before storyboarding: (1) who *specifically* is the
+audience and what do they already know; (2) the ONE main message they remember a week later; (3) the 3–5
+supporting points that carry it; (4) the evidence proving each; (5) the single call-to-action; (6) what is
+essential vs. expandable under time pressure. These feed the deck arc directly: message = thesis, points =
+beats, CTA = closing ask (→ `references/storytelling-and-decks.md`).
+
+**Assertion-evidence is the slide unit.** Each slide = one assertion (a complete claim, written as the
+title) + the visual evidence that proves it — never a topic label over a bullet list. *"User engagement rose
+43% after the redesign"* + a chart, not *"Engagement"* + three bullets. The body proves the headline; it
+never repeats it.
+
+**One concept per slide.** Working memory is small and the audience is also listening to you. If a slide
+needs two breaths to explain, it's two slides. Reveal sequential parts progressively — build order *is* the
+explanation — rather than dumping everything at once.
+
+**Spoken vs. shown — never both.** The slide and your mouth are two channels; redundancy wastes both. *Show*
+the assertion, the visual, the number, the next step; *say* the elaboration, the context, the interpretation,
+the story. Reading slides verbatim is the fastest way to lose a room.
+
+Full frameworks — the planning questions, the spoken/shown table, the **1–5 evaluation rubric**
+(audience-centered / visual clarity / cognitive load / accessibility), the implementation checklist, and the
+communication anti-patterns — live in `references/slide-design.md`. Score any draft against the rubric (≥ 4
+on each axis) before shipping.
+
+## Which pipeline? (decide before building)
+
+Pick once, up front — switching mid-build is expensive.
+
+| Question | → Markdown deck (Marp/Slidev) | → Native PPTX (python-pptx) |
+| --- | --- | --- |
+| Primary deliverable | A **stunning** deck; PDF is the hero, PPTX a bonus | An **editable `.pptx`** people will open and change |
+| Who edits after handoff | You / engineering, in Markdown + Git | Non-technical stakeholders, in PowerPoint/Keynote/Slides |
+| Visual ceiling | High — full CSS/HTML, web fonts, CSS grid, gradients, SVG | PowerPoint's box model; native charts/tables/SmartArt-lite |
+| Theming source | Design tokens → CSS theme (OKLCH, type scale, spacing) | Design tokens → `.pptx` theme (sRGB colors, theme fonts) |
+| Data viz | SVG / chart libs / images, full control | **Native, editable** PowerPoint charts (live in the file) |
+| Diffable / reviewable in Git | Yes (Markdown) | No (binary) |
+| Speaker notes | Yes (`<!-- notes -->` / `notes:`) | Yes (native notes pane) |
+| Version control of changes | Excellent | Poor (binary blobs) |
+
+**Default to the Markdown pipeline** for anything where "stunning" matters and the user is fine getting a
+PDF (+ image-based PPTX). **Choose python-pptx** the moment the user says "I need to edit it in
+PowerPoint", "the client edits the slides", "live charts", or "corporate template `.potx`". When unsure,
+ask one question: *"After I hand it over, will someone edit the slides in PowerPoint/Keynote, or is a
+polished PDF enough?"* You can also build in Markdown and additionally export a `--pptx` for handoff —
+just warn that Marp/Slidev PPTX slides are **images, not editable text** (a key gotcha, see exports).
+
+Deep recipes for each → `references/markdown-decks.md` and `references/pptx-python.md`.
+
+## Workflow
+
+1. **Ground in the brand study** (gate above). Pull voice, positioning, proof, audience.
+2. **Pick the deck arc** for the purpose (pitch / sales / product / keynote / investor / QBR) from `references/storytelling-and-decks.md`. Lock the one-sentence thesis the whole deck proves.
+3. **Write the slide-by-slide skeleton** — one assertion headline per slide + the proof it carries. This is a `marketing` collaboration: headlines are copy. Get the skeleton approved before designing pixels.
+4. **Pick the pipeline** (table above).
+5. **Build the theme from design tokens** — map the project's OKLCH palette, type scale, and spacing into a Marp/Slidev CSS theme or a python-pptx theme. (→ `references/slide-design.md`, `references/markdown-decks.md`, `references/pptx-python.md`)
+6. **Lay out the slides** against the visual system: grid, type scale for projection, data-viz best practices, imagery, contrast, restrained motion. (→ `references/slide-design.md`)
+7. **Produce presenter + leave-behind variants** if needed: presenter version is sparse (headline + visual, talk track in notes); leave-behind is self-explanatory (more on-slide text, appendix). (→ `references/storytelling-and-decks.md`)
+8. **Export** to PDF (vector, fonts embedded, 16:9) and/or editable PPTX; handle font-embedding and file-size gotchas. (→ `references/markdown-decks.md`, `references/pptx-python.md`)
+9. **Verify** with `scripts/verify.sh` (lint deck sources, dry export, import check) and the QA gate below.
+10. **Record deck conventions** in `docs/wiki/stack/presentations.md` (Project grounding, below).
+
+## Worked example — storyboard → theme → export (Markdown pipeline)
+
+One end-to-end pass, brand study already complete. Read this once and you rarely need to round-trip the
+references for a standard Markdown deck.
+
+**1. Storyboard the spine** (assertion headlines only — read top-to-bottom, they ARE the pitch; →
+`references/storytelling-and-decks.md`). Thesis: *"Onboarding v2 is why we can raise now."*
+
+```text
+1 Onboarding v2 cut churn 40%          (title)
+2 One in three users never finished setup   (problem — make it ache)
+3 We rebuilt the first run as one screen      (solution, one visual)
+4 Activation rose 28pts in six weeks          (proof — chart, one series)
+5 TAM is $12B, growing 24%/yr                 (market, stated assumptions)
+6 The ask: $2M to make this the default path  (CTA — one ask)
+```
+
+**2. Theme from design tokens.** Pull OKLCH palette + type scale from `docs/wiki/stack/design.md` into a
+Marp CSS theme (full theme → `references/markdown-decks.md`). The load-bearing move is mapping tokens to
+variables *once*, never hand-picking hex per slide:
+
+```css
+/* @theme brand — generated from docs/wiki/stack/design.md */
+@import url('https://fonts.googleapis.com/css2?family=Fraunces:wght@700&family=Inter:wght@400;600&display=swap');
+:root { --brand: oklch(0.62 0.19 264); --ink: oklch(0.18 0.03 264);
+        --surface: oklch(0.98 0.005 264); --accent: oklch(0.74 0.17 52); }
+section { background: var(--surface); color: var(--ink); font-family: Inter, sans-serif; font-size: 26px; }
+h1 { font-family: Fraunces, serif; font-size: 44px; } strong { color: var(--accent); }
+```
+
+Write the spine into `deck.md` with `marp: true`, `theme: brand`, `size: 16:9`, talk track in `<!-- notes -->`.
+
+**3. Export to PDF + PPTX** and verify fonts embed (gotchas → `references/markdown-decks.md`):
+
+```bash
+npx @marp-team/marp-cli@latest deck.md --theme ./theme.css --pdf --pdf-outlines --pdf-notes
+npx @marp-team/marp-cli@latest deck.md --theme ./theme.css --pptx     # image-per-slide; NOT editable text
+pdffonts deck.pdf            # every font must read 'emb yes'
+./scripts/verify.sh          # lint + dry export + QA checklist
+```
+
+If the client will *edit* the slides, this is the wrong pipeline — rebuild with python-pptx (→
+`references/pptx-python.md`), whose `build_deck.py` emits the same six slides as native editable shapes/charts.
+
+## Tooling & current versions (verified 2026-06)
+
+- **Marp** — `@marp-team/marp-cli`. Run pinned: `npx @marp-team/marp-cli@latest deck.md --pdf`. Exports HTML / PDF / PPTX / PNG / JPEG. Needs a Chromium-family browser (Chrome/Edge) or Firefox for PDF/PPTX/image export (v4 added Firefox via WebDriver BiDi as a fallback; Chrome/Edge are preferred and give the most faithful PDF). `--pptx` is image-per-slide; `--pptx-editable` is **experimental** and needs LibreOffice (`soffice`). `--notes` / a `.txt` output exports speaker notes. Node 18+. (→ `references/markdown-decks.md`)
+- **Slidev** — `@slidev/cli` (Vue-based). Scaffold `npm init slidev@latest`; dev `slidev`; export `slidev export` (PDF default; `--format pptx|png`). Export needs `playwright-chromium` installed in the project (`npx playwright install chromium` or `npm i -D playwright-chromium`). PPTX is image-per-slide; notes carry over per slide. Best for code-heavy / developer talks (live code, Monaco, Mermaid, Vue components). (→ `references/markdown-decks.md`)
+- **python-pptx** — `pip install python-pptx` (current major `1.x`, e.g. `1.0.x`). Pure Python, **no** Office/LibreOffice needed to write `.pptx`. Creates masters/layouts, text, tables, **native editable charts**, images, speaker notes. Cannot render to PDF itself — convert via LibreOffice `soffice --headless --convert-to pdf` or open in Office. (→ `references/pptx-python.md`)
+- **decktape / Playwright** — fallback HTML→PDF for any web deck (reveal.js, custom HTML) when Marp/Slidev export isn't available. (→ `references/markdown-decks.md`)
+
+Always pin/verify the version in the target project before generating (`marp --version`, `npx slidev --version`, `python -c "import pptx; print(pptx.__version__)"`). Tooling moves; re-check rather than trusting memory.
+
+## Slide copy (with `marketing`)
+
+The deck's words are conversion copy on a stage. Defer the *craft* to `../marketing/SKILL.md`; this skill
+enforces the deck-specific shape:
+
+- **Benefit-led, climbing feature → benefit → proof**, stopping at the rung the audience cares about. Specificity (a number, a mechanism, a receipt) beats adjectives — "2× faster" not "blazing fast".
+- **Two text densities, chosen deliberately:** *presenter* slides carry a headline + one visual + 0–3 support points, with the argument in the speaker notes / your mouth; *leave-behind* slides are self-contained because no one is narrating. Never ship a wall-of-text presenter slide.
+- **Voice from the brand study.** Headlines obey the do/don't word lists and tone samples. Ban-list words ("revolutionary", "seamless", "game-changer", "supercharge") are defects.
+
+## Visual system for slides (with `design`)
+
+The deck's pixels are the brand's design system projected at 3 metres. Defer the *system* to
+`../design/SKILL.md`; this skill enforces the deck-specific constraints (full depth →
+`references/slide-design.md`):
+
+- **Layout grid** built for 16:9: a 12-column grid, generous margins, one focal point per slide, consistent safe-area so nothing clips on a projector.
+- **Type scale for projection**, not laptop reading distance: display/headline/body/caption steps, body ≥ 24pt (≥ 28–32pt for talks), ~6 words/line and ~6 lines/slide as a ceiling, contrast ≥ 4.5:1. Never drop below the legibility floor to cram text — split the slide instead. If it doesn't read at 3 metres, it doesn't ship.
+- **Color from tokens, allocated for a room:** dark themes read better in dark rooms / on big screens, light themes for printed handouts and bright rooms. High contrast always; never rely on color alone to encode meaning.
+- **Data viz that makes one point:** one chart = one takeaway named in the headline; remove gridlines/clutter; label directly; pre-attentive emphasis (one highlighted bar/line) over rainbow palettes; never a 3-D pie. Every number traces to a source — mark gaps `[[NEEDS PROOF]]`, never fabricate one to fill a chart.
+- **Imagery** full-bleed and intentional (with a legibility scrim behind text), not stocky decoration; respect resolution so it doesn't pixelate on a 4K projector.
+- **Motion with restraint:** one transition family, ≤ 300ms, builds that reveal one idea at a time; reduced-motion honored in HTML pipelines. Animation explains sequence/state change, never just fills time.
+
+## Anti-patterns
+
+| Anti-pattern | Why it's wrong | Do instead |
+| --- | --- | --- |
+| Topic-label headlines ("Market", "Team") | Forces the audience to find the point | Assertion headline that states the point |
+| Wall of bullets / paragraphs on a slide | Audience reads instead of listening; nothing lands | One idea, ≤ ~6 lines; move detail to notes/appendix |
+| Reading the slides verbatim | The slide and the talk become redundant | Slide = the visual; you = the narration; notes = the script |
+| Tiny text to fit more | Unreadable from the back; signals filler | Split into more slides; raise the floor, not lower it |
+| Hand-picked hex per slide | Drifts off-brand, inconsistent | Map design tokens once into the theme |
+| Rainbow charts, 3-D pies, dual axes | Decoration over meaning; misleads | One highlighted series; direct labels; honest axes |
+| Generic stock photos + purple gradient | Reads as "AI deck"; no identity | Brand imagery + token palette + a real type pairing |
+| Every-element animation, slow transitions | Noise; tanks pacing; nausea | One fast transition family; builds that reveal meaning |
+| Exporting PPTX from Marp/Slidev and calling it "editable" | Slides are flattened images, not text | Use python-pptx when editability is required |
+| Fonts not embedded in the PDF | Renders with fallback fonts on other machines | Embed fonts; verify (see exports gotchas) |
+| Invented metrics to fill a chart | Destroys credibility in the room | Cite the source; mark gaps `[[NEEDS PROOF]]` |
+
+## Quick reference
+
+```bash
+# --- Markdown: Marp (PDF is the hero; PPTX = images) ---
+npx @marp-team/marp-cli@latest deck.md --theme ./theme.css --pdf            # vector PDF, fonts embedded
+npx @marp-team/marp-cli@latest deck.md --theme ./theme.css --pdf --pdf-outlines --pdf-notes
+npx @marp-team/marp-cli@latest deck.md --pptx                               # image-per-slide PPTX
+npx @marp-team/marp-cli@latest deck.md --pptx --pptx-editable               # experimental, needs soffice
+npx @marp-team/marp-cli@latest deck.md --notes notes.txt                    # speaker notes only
+
+# --- Markdown: Slidev (code-heavy talks) ---
+npm init slidev@latest                                                      # scaffold
+npx slidev                                                                  # dev server (localhost:3030)
+npx playwright install chromium                                             # one-time, for export
+npx slidev export                                                           # PDF (default)
+npx slidev export --format pptx                                             # image-per-slide PPTX
+
+# --- Native editable PPTX (python-pptx) ---
+pip install python-pptx
+python build_deck.py                                                        # your generator (see ref)
+soffice --headless --convert-to pdf deck.pptx                               # PPTX -> PDF via LibreOffice
+
+# --- Fallback: any HTML deck -> PDF ---
+npx decktape reveal http://localhost:8000 deck.pdf
+
+# --- Verify before shipping ---
+./scripts/verify.sh            # warn-by-default; lint sources + dry export + import check
+./scripts/verify.sh --strict   # gate CI (warnings become failures)
+```
+
+## Deck QA gate
+
+Run before claiming done — each line is a defect if unchecked. `scripts/verify.sh` automates the
+mechanical subset.
+
+- [ ] Brand study located, complete, and cited (which articles grounded the deck).
+- [ ] One idea per slide; every slide titled with an assertion headline, not a topic label.
+- [ ] Body text ≥ 24pt (≥ 28–32pt for a talk); ≤ ~6 words/line, ≤ ~6 lines/slide; contrast ≥ 4.5:1.
+- [ ] Colors / type / spacing come from design tokens, not hand-picked per slide.
+- [ ] Deck follows a deliberate arc with a single thesis; opens with a hook, closes with the ask/CTA.
+- [ ] Every number traces to a source; gaps marked `[[NEEDS PROOF]]`, none invented.
+- [ ] Charts each make one point, named in the headline; no 3-D/rainbow/dual-axis clutter.
+- [ ] Motion is one restrained family (≤ 300ms); reduced-motion honored (HTML); builds reveal meaning.
+- [ ] Presenter vs leave-behind variant chosen deliberately; presenter notes hold the talk track.
+- [ ] 16:9; PDF is vector with fonts embedded; PPTX opens clean in PowerPoint/Keynote/Slides.
+- [ ] PPTX editability matches the promise (python-pptx if "editable", not flattened Marp/Slidev images).
+- [ ] File size sane (compressed images, subsetted fonts); ban-list words absent from copy.
+- [ ] Scored ≥ 4/5 on each axis of the design rubric — audience-centered, visual clarity, cognitive load, accessibility (→ `references/slide-design.md`, "Diagnostic rubric").
+
+## Project grounding (docs)
+
+In a project with a `docs/` layer (the [`harness`](../harness/SKILL.md) Karpathy wiki), read
+`docs/wiki/stack/presentations.md` first and stay consistent with it; create or refresh it when
+missing or stale, and index it in `docs/wiki/index.md`. It records this project's real choices: the
+chosen pipeline (Marp / Slidev / python-pptx) and why, the theme file path and how it maps the design
+tokens, the standard deck arc(s), export commands and the canonical output, the presenter-vs-leave-behind
+convention, and font-embedding / asset-location notes.
+
+The deck theme is downstream of the design tokens: always reconcile that article with
+`docs/wiki/stack/design.md` so the deck and the product share one palette and type system.
+
+No `docs/` layer? Skip silently (optionally suggest `harness`). Unlike the brand study, deck
+conventions are *recorded, not gated* — never block the task on this.

package/skills/presentations/evals/README.md

@@ -0,0 +1,61 @@
+# Eval harness — `presentations`
+
+These evals are run by an **agent harness** (an LLM agent with the skill catalog
+available), not a pure shell script. There is no automated grader; a human or a
+judge-agent scores against the rubrics in `cases.yaml`. Two things are measured:
+**triggering** (does the right skill fire?) and **capability** (does the skill make
+the answer measurably better?).
+
+## Setup
+
+- One agent, the full skill catalog discoverable by description only (`fastapi`,
+  `nextjs`, `go`, `postgresdb`, `flutter`, `design`, `marketing`, `presentations`,
+  `course-storytelling`, `building-agents`, `secure-coding`, `deployment`, `harness`,
+  `init`). The agent must choose which skill(s) to load — do not pre-load any.
+- Run each prompt in a fresh session (no memory carryover between trials).
+
+## 1. Triggering
+
+For every prompt in `should_trigger` and `should_not_trigger`:
+
+1. Start a clean session with only skill **descriptions** visible.
+2. Paste the prompt verbatim. Observe which skill the agent loads/invokes first.
+3. Run **3–5 trials** per prompt (phrasing is fuzzy; repetition catches flakiness).
+
+Scoring per prompt:
+
+- `should_trigger` → **pass** if `presentations` is invoked in a strong majority of
+  trials (>= 4/5).
+- `should_not_trigger` → **pass** if `presentations` is **not** invoked, AND ideally
+  the agent routes to the `route_to` sibling (or asks/declines when `route_to: none`).
+  Loading `presentations` here is a **fail** (false positive).
+
+**Pass bar:** >= 90% trigger accuracy across all cases (false positives are weighted
+heavily — a near-miss that fires is worse than a true-positive that misses once).
+
+## 2. Capability (with vs without the skill)
+
+For each `capability` scenario, run the same prompt twice:
+
+- **WITHOUT** — agent answers with the skill **not** loaded (baseline).
+- **WITH** — agent answers with `presentations` loaded.
+
+Grade both outputs against that scenario's `must_include` checklist (one point per
+item covered substantively, not just name-dropped). Run 2–3 trials each side and
+average to reduce variance.
+
+**Pass bar:**
+
+- WITH covers **>= 80%** of `must_include` points.
+- WITH beats WITHOUT by a clear margin (the brand-grounding gate, assertion-evidence
+  discipline, token-driven theming, legibility floors, `[[NEEDS PROOF]]`, and
+  export verification should be largely absent or vague WITHOUT the skill).
+- If WITHOUT already satisfies the rubric, the case is too generic — tighten it.
+
+## Notes
+
+- Judge on substance, not keywords: "grounds copy in the brand study" must show as
+  an actual gate (locate → check → interview/cite), not a passing mention.
+- Keep prompts verbatim; edits to wording invalidate a trial.
+- Log per-trial which skill fired and which rubric points were hit, so failures are
+  diagnosable (mis-route vs. weak capability).

package/skills/presentations/evals/cases.yaml

@@ -0,0 +1,56 @@
+skill: presentations
+
+should_trigger:
+  - prompt: "Build me an investor pitch deck for our seed round — 12 slides, exported to PDF."
+    why: "Core use case: an investor/pitch deck with a defined slide count and PDF export. The skill owns deck structure, theme, and export."
+  - prompt: "Turn this product memo into slides I can present at the all-hands next week."
+    why: "Doc-to-deck conversion for an internal all-hands is explicitly in scope ('Turning an existing doc, memo, or notes into a slide-by-slide deck')."
+  - prompt: "I need an editable .pptx of our Q3 board review that the CFO can change in PowerPoint."
+    why: "Native editable PPTX hand-off for a non-technical stakeholder is the python-pptx pipeline this skill orchestrates; QBR/board review is a named deck type."
+  - prompt: "Make my conference talk look incredible — it's code-heavy and I want live syntax highlighting on the slides."
+    why: "A keynote/conference talk with code-heavy content is in scope; the skill routes this to the Slidev pipeline. User never names a tool but describes a deck."
+  - prompt: "Our sales slides look like a generic AI template. Re-skin them to match our brand and tighten the story."
+    why: "Re-skinning/restructuring an existing deck to be on-brand and less generic is an explicit Use-when bullet; spans the brand-grounding gate plus theme-from-tokens."
+  - prompt: "Can you whip up a one-page leave-behind exec summary we can hand out after the demo?"
+    why: "A leave-behind one-pager / executive summary is an explicit deliverable this skill produces, including the presenter-vs-leave-behind variant logic."
+  - prompt: "Storyboard a keynote that opens with a hook, has one idea per slide, and ends with a clear ask."
+    why: "Deck arc, assertion-headline storyboarding, one-idea-per-slide, and a closing CTA are the skill's core narrative/design responsibilities, even without naming PDF/PPTX."
+
+should_not_trigger:
+  - prompt: "Write the headline, value prop, and three benefit bullets for our landing hero — just the copy."
+    route_to: "marketing"
+    why: "Writing the words/copy in isolation (no deck layout or export) is explicitly delegated to marketing; this skill consumes that copy, it does not author it."
+  - prompt: "Define our brand's OKLCH color palette, type pairing, and spacing scale as design tokens."
+    route_to: "design"
+    why: "Choosing the visual tokens (palette, type, spacing, motion personality) is design's job; presentations only maps existing tokens into a slide theme."
+  - prompt: "Build a scroll-driven animated landing page for our product launch in Next.js."
+    route_to: "nextjs"
+    why: "A web-native, scroll/animation-heavy landing page is not a deck; the build belongs to design + nextjs, per the explicit Do-NOT-use bullet."
+  - prompt: "Produce a 60-second animated explainer video of how our product works, with motion graphics."
+    route_to: "none"
+    why: "A video/motion explainer as the deliverable is explicitly out of scope — this skill uses motion only as restrained slide transitions, never as the medium. No catalog sibling owns video."
+  - prompt: "Build the financial model spreadsheet — revenue projections, cohort retention, and a 5-year P&L."
+    route_to: "none"
+    why: "A live financial model/spreadsheet (the numbers engine) is explicitly excluded; this skill renders slides and cites the model as source of truth. No catalog sibling owns modeling."
+  - prompt: "Generate a Postgres query to compute monthly churn so we have a number for the deck."
+    route_to: "postgresdb"
+    why: "A pure-SQL/data task belongs to postgresdb; the deck only consumes the resulting number. The mention of 'deck' is incidental, not a deck-building request."
+
+capability:
+  - scenario: "User asks: 'Build a 6-slide seed pitch deck. Thesis: onboarding v2 is why we can raise now. We have churn and activation numbers. Export to PDF, and we may need an editable PPTX for the CFO.'"
+    must_include:
+      - "Runs the brand-grounding gate FIRST — locate brand study in docs/wiki/brand/ via root CLAUDE.md; STOP and interview if absent/incomplete; never fabricate voice; cite which brand articles grounded the deck."
+      - "Storyboards an assertion-headline spine (one assertion per slide, e.g. 'Onboarding v2 cut churn 40%'), not topic labels, with one idea per slide and a single locked thesis."
+      - "Picks a pipeline deliberately: Markdown (Marp/Slidev) when a stunning PDF is the hero vs python-pptx when the CFO must edit in PowerPoint; asks the disambiguating question or builds Markdown + warns Marp/Slidev PPTX is image-per-slide, not editable text."
+      - "Themes from design tokens (OKLCH palette, type scale, spacing from docs/wiki/stack/design.md) mapped once into the theme — no hand-picked hex per slide."
+      - "Enforces projection legibility: body >= 24pt (28-32pt for a talk), ~6 words/line, ~6 lines/slide ceiling, contrast >= 4.5:1, 16:9."
+      - "Requires every metric to trace to a source and marks gaps [[NEEDS PROOF]] — never invents numbers to fill a chart."
+      - "Exports verified: PDF vector with fonts embedded (e.g. pdffonts shows 'emb yes'), runs scripts/verify.sh / the Deck QA gate before claiming done."
+  - scenario: "User asks: 'Make my conference talk slides legible and not death-by-bullets. There's a chart showing engagement went up after a redesign.'"
+    must_include:
+      - "Applies assertion-evidence: the slide title states the claim ('User engagement rose 43% after the redesign') and the chart is the evidence — not a 'Engagement' label over bullets."
+      - "Separates spoken vs shown: minimal on-slide text, talk track / elaboration moved to speaker notes; warns against reading slides verbatim."
+      - "Sets talk-grade type: body ~28-32pt, ~6 words/line and ~6 lines/slide as a ceiling; splits a crowded slide into more slides rather than shrinking text."
+      - "Designs the chart to make ONE point named in the headline: removes gridlines/clutter, direct labels, one pre-attentive highlighted series; bans 3-D pies, rainbow palettes, dual axes."
+      - "Uses restrained motion: one transition family <= 300ms, builds that reveal one idea at a time, honors reduced-motion in HTML pipelines — never animation as decoration."
+      - "Bans filler/hype words (revolutionary, seamless, game-changer, supercharge) and scores the draft >= 4/5 on the design rubric axes (audience-centered, visual clarity, cognitive load, accessibility) before shipping."

package/skills/presentations/references/brand-grounding.md

@@ -0,0 +1,160 @@
+# Brand Grounding for Decks — Checklist, Question Script & Persistence
+
+A deck is the brand on stage, so it grounds in the **same** brand study the `marketing` and `design` skills
+use — it does not fork it. This file holds the **completeness checklist** (the shared brand dimensions plus the
+**deck-specific** ones), the **question script** (how to interview, batched), and the **persistence format**
+(how to write it into `docs` and link it from `CLAUDE.md`). The runtime hard-STOP that invokes this lives in
+`SKILL.md` under "Brand grounding (read this first)".
+
+## Where the brand study lives
+
+Following the `harness` Karpathy-wiki convention (identical to `marketing`/`design`):
+
+```text
+docs/
+├── raw/brand/         ← immutable: voice samples + raw inputs pasted verbatim by the user
+└── wiki/brand/        ← compiled brand study, one article per dimension
+    ├── index.md            ← positioning one-liner + links to every dimension article
+    ├── audience.md         ← ICP, pains, desires (+ deck audience & setting, below)
+    ├── value-proposition.md← value prop, differentiation
+    ├── voice.md            ← tone, do/don't word lists, links to raw voice samples
+    ├── proof.md            ← credibility, metrics, customers, guarantees
+    ├── offers.md           ← offers + primary/secondary CTA (the deck's ASK)
+    └── channels.md         ← channels (incl. how decks are delivered/presented)
+```
+
+Deck-specific facts (purpose, length, presenter-vs-leave-behind, must-include slides) are **per-deck**, not
+brand-permanent — capture them at the start of each deck and record the *convention* in
+`docs/wiki/stack/presentations.md` (see SKILL.md "Project grounding"). The brand-permanent dimensions are
+shared with marketing/design and live in `wiki/brand/`.
+
+## Completeness checklist
+
+**COMPLETE** only when every dimension is filled with real, specific content. Any empty / placeholder / "TBD" =
+**INCOMPLETE** = hard STOP, interview the user.
+
+### Shared brand dimensions (same as marketing/design — reuse if already filled)
+
+- [ ] **1. Brand name & one-line positioning** — "[Brand] helps [audience] [outcome] by [mechanism]." No jargon.
+- [ ] **2. ICP / audience, pains & desires** — who exactly; top 2–3 pains in their words; top 2–3 desires; today's alternatives.
+- [ ] **3. Value proposition & differentiation** — core benefit in one sentence; what's different; the one thing only this brand can say.
+- [ ] **4. Tone & voice** — 3–5 tone adjectives; a **do-words** and **don't-words** list; **3–5 voice samples pasted from the user's real writing** (mandatory — never infer a voice).
+- [ ] **5. Proof & credibility** — real metrics, named customers/logos, quantified testimonials, certifications, guarantees, traction. Mark aspirational items `[[NEEDS PROOF]]`.
+- [ ] **6. Offers & primary CTA** — what the audience is asked to do; the primary CTA verb; any secondary CTA; pricing if relevant. (This is the deck's **ask**.)
+- [ ] **7. Channels** — where the brand publishes / how decks are delivered (in-person, sent PDF, webinar).
+
+### Deck-specific dimensions (capture per deck, before building)
+
+- [ ] **8. Deck purpose & type** — pitch / sales / product launch / keynote / investor / board-QBR / training / internal. What single outcome must this deck achieve?
+- [ ] **9. Audience & setting** — who's in the room (role, knowledge level, skepticism); the setting (live on stage / Zoom screen-share / sent as a PDF to read alone); the time slot (5-min pitch vs 45-min talk).
+- [ ] **10. Length** — target slide count / time. (5–10 short, 10–20 medium, 20+ long; keynotes time-boxed.)
+- [ ] **11. Presenter vs leave-behind** — are you narrating it live (sparse slides + speaker notes) or is it read unattended (self-contained), or **both** (build presenter, derive leave-behind)?
+- [ ] **12. Must-include & must-avoid** — mandatory slides/sections (a specific metric, a partner logo, a legal disclaimer, a required template), and anything that must NOT appear.
+- [ ] **13. Visual & format constraints** — pipeline preference if any (Marp/Slidev/editable-PPTX); a corporate template (`.potx`) to honor; dark vs light; aspect ratio (default 16:9); the deck theme must reconcile with the design tokens in `docs/wiki/stack/design.md`.
+
+## Question script (ask in batches, never all at once)
+
+Send a batch, wait, persist, send the next. **Skip questions the located study already answers** — only fill
+gaps. Stop the moment every dimension is complete.
+
+### Batch A — purpose, audience, format (deck-specific, ask first)
+
+```text
+1. What kind of deck is this (pitch / sales / product launch / keynote / investor / board-QBR /
+   training / internal), and in one sentence: what's the ONE outcome it must achieve?
+2. Who's the audience and what's the setting — live on stage, a Zoom screen-share, or a PDF they
+   read alone? How skeptical/knowledgeable are they, and how long is your slot?
+3. Roughly how many slides / minutes?
+4. Will you narrate it live (sparse slides, I write your talk track into speaker notes), is it sent
+   to read unattended (self-contained), or both?
+5. Any must-include slides or sections (a specific metric, a partner logo, a required template),
+   and anything that must NOT appear?
+6. Any format constraints — a corporate template (.potx) to match, dark vs light, do you need an
+   EDITABLE PowerPoint at the end or is a polished PDF enough?
+```
+
+### Batch B — brand identity & audience (skip if `wiki/brand/` already has it)
+
+```text
+7. Brand/product name, and in one sentence: who is it for and what outcome does it deliver?
+8. Describe the ideal customer/decision-maker precisely — role, context, sophistication.
+9. In THEIR words, the top 2–3 pains that make them look for something like you?
+10. The top 2–3 things they actually want? And what do they use today instead of you?
+```
+
+### Batch C — value & proof (skip if already filled)
+
+```text
+11. In one sentence, the core benefit — no feature list, no jargon?
+12. What makes you genuinely different from those alternatives, and what's the one thing only YOU
+    can credibly claim?
+13. What proof can you stand behind RIGHT NOW (metrics, named customers, quantified testimonials,
+    certifications)? If something's aspirational, say so — I'll mark it [[NEEDS PROOF]], not state it as fact.
+```
+
+### Batch D — voice & the ask (voice samples mandatory; skip if already filled)
+
+```text
+14. 3–5 adjectives for how the deck should sound, plus a DO list and a DON'T list of words/phrases.
+15. Paste 3–5 samples of your REAL writing (posts, emails, prior decks/copy) so I match your voice
+    instead of inventing a generic one — I'll save them verbatim under docs/raw/brand/.
+16. What's the single action you want the audience to take by the last slide (the ask)? Any
+    secondary action?
+```
+
+If the user can't answer a dimension, it stays INCOMPLETE — note the gap, keep the STOP for that dimension,
+and offer to draft a hypothesis they confirm rather than fabricating. Voice samples are non-negotiable: no
+samples → no voice → don't write the headlines.
+
+## Persistence format
+
+When a dimension is newly answered, write it as a wiki article under `docs/wiki/brand/` (shared dimensions)
+and record deck conventions under `docs/wiki/stack/presentations.md`. Save any pasted raw text verbatim under
+`docs/raw/brand/` and link it from the article's `> Raw:` line. Article format:
+
+```markdown
+# Voice
+
+> Updated: 2026-06-01 · Source: user interview + raw/brand/voice-sample-1.md
+
+Tone: direct, dry, concrete. ...
+
+## Do words
+ship, mechanism, receipts, ...
+
+## Don't words
+revolutionary, seamless, game-changer, supercharge, ...
+
+> Raw: see [raw/brand/voice-sample-1.md](../../raw/brand/voice-sample-1.md)
+```
+
+Then add/refresh the link in the root `CLAUDE.md` (additive — never delete existing sections). Insert (or
+update) under a `## Brand & voice` heading:
+
+```markdown
+## Brand & voice
+
+Brand study (source of truth for all marketing, design, and deck copy):
+`docs/wiki/brand/` — see `index.md` for positioning and links to every dimension.
+Raw voice samples & inputs: `docs/raw/brand/`.
+```
+
+And ensure the deck conventions are indexed from the `## Knowledge map` section (create if absent):
+
+```markdown
+## Knowledge map
+
+- Deck conventions: `docs/wiki/stack/presentations.md`
+- Design tokens: `docs/wiki/stack/design.md`
+```
+
+Update `wiki/index.md` and `wiki/log.md` per the harness convention. Create any missing directories/files
+(`CLAUDE.md` included) — additive only.
+
+## Why this gate exists
+
+Generic decks are the failure mode. A deck with no brand behind it defaults to the AI-median look and voice —
+the exact "another AI deck" the audience tunes out. Grounding every headline in the voice samples and every
+visual in the design tokens is what makes a deck unmistakably *this* brand. Cite the articles you used in the
+deliverable (e.g. "narrative grounded in `value-proposition.md`, voice in `voice.md`, palette in
+`stack/design.md`") so the user can trace every choice back to the study.