Build a Fundraise Deck Workspace

What This Blueprint Is

A canonical playbook for building a private, gated, two-surface fundraise/teaser deck workspace in the astro-knots monorepo. The canonical reference implementation is sites/calmstorm-decks. Every section below points at concrete files in that site so a new build can lift the pattern verbatim and adapt it.

Audience: a Claude Code session (or human collaborator) initializing a new deck site for a different client. The user will have:

• A folder of source slides as PDFs and/or rough notes.
• A brand identity (colors, typography, voice) — at minimum, a wordmark and a couple of accent colors.
• A list of slide titles and an idea of what’s on each slide.
• An audience that is small, named, and trusted (LPs, partners, prospects) — not the public web.

This spec assumes the parent astro-knots/CLAUDE.md has been read and understood. It does not repeat material from there (workspace structure, LFM consumption pattern, two-tier CSS tokens, design-system + brand-kit maintenance motion, etc.). It captures only what’s specific to the fundraise-deck shape.

The Two-Surface Architecture (the most important decision)

A fundraise deck workspace serves two distinct audiences, and trying to serve both with one surface is the single most expensive mistake to avoid.

Surface A — Slide-by-slide variant chooser (design research)

URLs: /{slug} (chooser) and /drafts/{slug}/{slug}-vN (individual variant).

Audience: the build team. Granular A/B comparison of single-slide layouts. Each slug is one slide; each vN is one alternative composition for that slide. Used during early design exploration and whenever a single slide needs reinvention.

Surface B — Scroll-deck (stakeholder reading)

URLs: /thesis, /thesis/version-2, /thesis/version-N.

Audience: stakeholders, prospective LPs, partners. The full deck composed end-to-end as a single scrollable page with scroll-snap navigation. This is the deliverable that gets shared. Each version-N is an entire seventeen-slide composition with its own design voice.

Why both

• The slide-by-slide surface is where compositional risk gets explored cheaply, one slide at a time. Without it, every layout decision contaminates seventeen other decisions.
• The scroll-deck surface is the actual stakeholder artifact. Variant choosers are research scaffolding; the scroll deck is the product.
• The two surfaces share the same source data, the same theme tokens, the same chrome. Only the composition strategy differs.

This was the most expensive lesson of the calmstorm-decks build week. The first cut shipped only the slide-by-slide surface, on the assumption that the deck would be assembled later from the chosen variants. In practice, clients want the assembled artifact much sooner — and assembling it turned out to be its own first-class design problem deserving its own surface and its own variants. Plan for both from the start.

Build Sequence

Ordered so each step’s output is the input the next step needs. Skipping ahead works but costs back-and-forth.

Step 1 — Establish source content first (context-v/narratives/)

For every slide, write a markdown file at context-v/narratives/{NN}-{slug}.md that contains:

• YAML frontmatter with the slide’s structured fields (eyebrow, headline, subhead, key facts, etc.)
• A prose section: “What this slide is”, “Why it’s here”, “What’s most important to surface”, “Visual hierarchy I’d suggest”.

The reason this comes first: section components compose against this copy later. Without the narrative files, layout iteration is contaminated by copy iteration, and both move slower. Sharpen the words before you touch the layout.

Reference: sites/calmstorm-decks/context-v/narratives/01-disclaimer-confidential.md through 17-fund-terms.md.

Step 2 — Establish design tokens (DESIGN.md + src/styles/theme.css)

DESIGN.md at the site root documents the brand vocabulary in prose. Then src/styles/theme.css implements it in two tiers per the astro-knots CSS convention (full spec in parent CLAUDE.md → CSS Token Convention):

• Tier 1 — named palette: --color__blue-azure, --font__lato, etc.
• Tier 2 — semantic roles: --color-primary, --color-on-surface-strong, --font-heading-1, etc.

theme.css should also export the typographic utility classes that section components compose: .eyebrow, .headline, .section-title, .subtitle, .statement, .stat-large, .badge, .card, plus the .slide structure and reveal-animation helpers.

Reference: sites/calmstorm-decks/src/styles/theme.css. Light-mode only; no dark-mode work for v0.1.0.

Step 3 — Initialize the Astro project

Standard astro-knots site setup (per parent CLAUDE.md “Adding a New Client Site”). Confirm:

• output: "static"
• adapter: vercel()
• Astro 6 stable Fonts API entry for the display family
• vite: { plugins: [tailwind()] } for Tailwind 4
• A site URL placeholder in astro.config.mjs (env-overridable via SITE_URL) — needed for absolute OG/canonical URLs in Step 8

Add the standard .npmrc for workspace-standalone behavior.

Reference: sites/calmstorm-decks/astro.config.mjs.

Step 4 — Build foundational chrome (src/components/basics/)

Per the astro-knots convention, headers / footers / nav primitives / metadata components live under components/basics/. Build these four before anything else:

• DeckHeader.astro — thin top header with a 3-column grid (wordmark · navigation · meta). Includes inline <GateScript /> so every page that uses the header is gated automatically. The center nav surfaces the workspace’s first-class surfaces (TOC · Scroll · Changelog) with active-state highlighting.
• DeckNav.astro — fixed bottom-right counter + ‹ › nav, plus an optional ← / → keyboard handler. Presentational; takes counter, prev, next, cycling props. Used by SlideLayout (variant cycling) and by every scroll-deck page (variant cycling between scroll decks).
• MetaTags.astro — comprehensive <head> metadata: title, description, robots, canonical, full OpenGraph block (image + width
  • height + alt + secure_url + type), Twitter card. Defaults to noindex, nofollow, noarchive, nosnippet.
• GateScript.astro — synchronous inline <script is:inline> that redirects to / when the unlock flag is missing from localStorage and the current path isn’t already the cover. Imported once by DeckHeader.astro so it propagates to every gated page.

Reference: sites/calmstorm-decks/src/components/basics/.

Step 5 — Build the first scroll-deck composition (/thesis)

Stand up src/layouts/PageAsDeckWrapper.astro (scroll-snap container with keyboard nav, double-click nav, section indicator, reveal-on-intersect) and the seventeen section components at src/layouts/sections/teaser/T01-…T17-…astro. The page entry at src/pages/thesis/index.astro imports them all and drops them into the wrapper.

Each section component composes the theme.css vocabulary against the narrative content. Section components add their own scoped styles only for unique flourishes. The base look comes from theme.css.

Reference:

• sites/calmstorm-decks/src/layouts/PageAsDeckWrapper.astro
• sites/calmstorm-decks/src/layouts/sections/teaser/
• sites/calmstorm-decks/src/pages/thesis/index.astro

The wrapper hard-codes height: 100vh by default but accepts a CSS variable override (--deck-height) so pages with chrome above the deck can shrink it: e.g. body.thesis-body { --deck-height: calc(100vh - 3rem); } when DeckHeader is present.

Step 6 — Add a second scroll-deck variant (/thesis/version-2)

Copy src/layouts/sections/teaser/ to src/layouts/sections/teaser-v2/ and reinvent every section’s layout while preserving content. The constraint: each slide in v2 must make a substantially different layout choice from v1 — different information architecture, not a re-skin.

Then add the variant to the registry (Step 7) and create src/pages/thesis/version-2.astro.

This is the moment where the value of the two-surface architecture becomes obvious. Adding a second compositional voice costs ~one Claude session because all the foundation work (narratives, tokens, wrapper, chrome) is already paid for.

Step 7 — Wire the scroll-deck registry (src/lib/scroll-decks.ts)

A registry is the single source of truth for what scroll-deck variants exist. The header pill, the bottom-right DeckNav, the keyboard handler, and the index-page CTA all read from it. Adding a new scroll deck is one line of TypeScript:

export const SCROLL_DECKS: ScrollDeck[] = [
  { href: "/thesis",           label: "v1 · baseline",  variantNumber: 1 },
  { href: "/thesis/version-2", label: "v2 · alternate", variantNumber: 2 },
];

Plus exported helpers isScrollDeckPath() and getScrollDeckCycle() — both used by DeckHeader and the per-page DeckNav.

Reference: sites/calmstorm-decks/src/lib/scroll-decks.ts.

Step 8 — Wire OpenGraph + SEO

Three pieces:

1. src/lib/seo.ts — central registry of per-page metadata. Constants: SITE_NAME, DEFAULT_OG_IMAGE and its dimensions, alt text, TITLE_SUFFIX. Per-page entries: SLIDE_SEO[slug], SCROLL_DECK_SEO[href], STATIC_SEO.root, STATIC_SEO.changelogIndex.
2. src/components/basics/MetaTags.astro — already created in Step 4. Drop into the <head> of every page. Reads constants from seo.ts and overrides via props.
3. astro.config.mjs — set site URL (env-overridable) so absolute URLs compose correctly.

Plus one shared OG image at public/og_image__{ClientName}-Deck-{Fund}.png for v1. Per-page custom images can be added later via the ogImage prop on MetaTags.

References:

• sites/calmstorm-decks/src/lib/seo.ts
• sites/calmstorm-decks/src/components/basics/MetaTags.astro
• sites/calmstorm-decks/astro.config.mjs

Step 9 — Add the polite access gate

Five files. Together they establish a closed-by-default posture for the workspace before it goes anywhere meaningful audiences can find it.

1. public/robots.txt — Disallow: / for compliant crawlers.
2. vercel.json — global X-Robots-Tag: noindex, nofollow, noarchive, nosnippet header for crawlers that ignore robots.txt but parse headers.
3. .env — PUBLIC_DECK_CODE=<the-shared-secret> (gitignored). .env.example committed as the template.
4. src/lib/gate.ts — three constants: GATE_STORAGE_KEY, GATE_UNLOCKED_CLASS, GATE_COVER_PATH.
5. src/components/basics/GateScript.astro — already created in Step 4.

The cover at / is restructured into two panes (.cover-pane + .menu-pane) with an inline <script is:inline> in <head> that synchronously adds html.cs-unlocked if the localStorage flag is set, so the wrong audience never sees a flash of the wrong pane.

Be honest about what this is: a politeness gate, not a security boundary. The deck HTML still ships statically; anyone with curl and a known URL bypasses everything. Acceptable for a brand-new URL with no inbound links. When the audience grows or the URL leaks, escalate to Astro middleware + Vercel SSR + cookie verification (option C in the exploration doc below).

References:

• sites/calmstorm-decks/src/lib/gate.ts
• sites/calmstorm-decks/src/components/basics/GateScript.astro
• sites/calmstorm-decks/src/pages/index.astro (cover + menu structure)
• sites/calmstorm-decks/context-v/explorations/Gate-Sensitive-Information-with-Simple-Code.md (full threat model + path to a real gate when needed)

Step 10 — Add the changelog surface

context-v/changelogs/{YYYY-MM-DD}_{NN}.md is the canonical location. Each entry has structured frontmatter (title, lede, dates, at_semantic_version, status, category, tags, authors, augmented_with) and a body with ## Why Care?, ## What Was Built, ## Open Items, ## Reference sections.

Render at:

• /changelog — list page with date · version · status · linked title · lede · collapsible “Why Care?” excerpt rendered through LFM.
• /changelog/[slug] — detail page with full body via LFM, prev/next neighbor links.

Loader at src/lib/changelog.ts globs the markdown files at build, parses frontmatter via gray-matter, derives the canonical date from the filename prefix, and slices the “Why Care?” section out of the body.

References:

• sites/calmstorm-decks/src/lib/changelog.ts
• sites/calmstorm-decks/src/pages/changelog/index.astro
• sites/calmstorm-decks/src/pages/changelog/[slug].astro
• sites/calmstorm-decks/context-v/changelogs/2026-05-01_01.md (template)

LFM components copied per the parent CLAUDE.md “Implementing @lossless-group/lfm” recipe — AstroMarkdown.astro, Callout.astro, CodeBlock.astro, MarkdownImage.astro go under src/components/markdown/.

Directory Conventions

sites/{client}-decks/
├── astro.config.mjs                  # site URL (env-overridable), Fonts API, Tailwind
├── package.json                      # rich metadata — see "package.json metadata" below
├── public/
│   ├── og_image__*.png               # shared social-preview image
│   └── robots.txt                    # Disallow: /
├── vercel.json                       # X-Robots-Tag header
├── .env                              # PUBLIC_DECK_CODE (gitignored)
├── .env.example                      # template, committed
├── DESIGN.md                         # brand vocabulary in prose
├── context-v/
│   ├── narratives/                   # one .md per slide — source content
│   ├── specs/                        # site-specific specs (this file goes here)
│   ├── explorations/                 # threat models, design explorations
│   ├── changelogs/                   # YYYY-MM-DD_NN.md build notes
│   └── extra/private/                # source PDFs, gitignored
└── src/
    ├── styles/
    │   ├── global.css                # Tailwind import
    │   └── theme.css                 # token cascade + utility classes
    ├── lib/
    │   ├── slides.ts                 # slide registry (slug → number, title)
    │   ├── scroll-decks.ts           # scroll deck variant registry
    │   ├── seo.ts                    # per-page metadata registry
    │   ├── gate.ts                   # gate constants
    │   └── changelog.ts              # changelog loader
    ├── data/                         # shared content data (team, pillars, etc.)
    ├── components/
    │   ├── basics/                   # DeckHeader, DeckNav, MetaTags, GateScript
    │   └── markdown/                 # AstroMarkdown, Callout, CodeBlock, MarkdownImage
    ├── layouts/
    │   ├── PageAsDeckWrapper.astro   # scroll-snap deck container
    │   ├── SlideLayout.astro         # 16:9 letterbox + chrome for slide-by-slide
    │   └── sections/
    │       ├── teaser/               # v1 scroll-deck sections (T01-…T17-)
    │       ├── teaser-v2/            # v2 scroll-deck sections
    │       └── teaser-vN/            # additional variants
    └── pages/
        ├── index.astro               # cover + menu (gated)
        ├── thesis/
        │   ├── index.astro           # /thesis (v1 scroll deck)
        │   ├── version-2.astro       # /thesis/version-2
        │   └── version-N.astro       # additional scroll variants
        ├── {slug}/index.astro        # variant chooser per slide
        ├── drafts/{slug}/{slug}-vN.astro  # individual slide variants
        └── changelog/
            ├── index.astro           # changelog list
            └── [slug].astro          # changelog detail

The Narrative-Driven Composition Recipe

The single biggest cost-collapse in this workspace pattern. Four ingredients that, together, let you generate new full-deck variants in a single session each instead of weeks each.

1. Pre-extracted, pre-sharpened source copy in context-v/narratives/{n}-{slug}.md. Done once. Section components compose against this copy — they don’t extract it from PDFs at build time and they don’t iterate on the copy alongside the layout.
2. A defined typography + token vocabulary in theme.css. The utility classes (.eyebrow, .headline, .section-title, .subtitle, .statement, .stat-large, .badge, .card, etc.) are the building blocks. Section components compose them; they don’t reinvent typography per slide.
3. A single rendering primitive — PageAsDeckWrapper.astro with scroll-snap. Each section gets a <section class="slide"> slot; the wrapper handles the rest.
4. @lossless-group/lfm for any markdown surface (changelogs, eventual narrative-driven slides). Don’t roll a markdown renderer.

Together these are what allowed v2 (17 alternate-layout sections, ~3,400 lines) to ship in a single Claude session, and v3 (editorial print magazine voice, also 17 sections) to do the same. The marginal cost of a new full-deck variant approaches one session of work.

The lesson is portable: AI-assisted design and authoring works much better when the model is composing against pre-decided vocabulary and pre-written content, rather than inventing both alongside the layout.

Pattern Catalog

Each entry: what it does, where the canonical implementation lives, and the one thing future builders most often get wrong.

PageAsDeckWrapper

Scroll-snap container with keyboard nav (↑ / ↓ / Home / End / Page Up / Page Down), double-click nav (upper half = prev, lower half = next), fixed section indicator, and IntersectionObserver reveal-on-intersect for .reveal-item children with per-item --delay.

• Reference: src/layouts/PageAsDeckWrapper.astro
• Common mistake: hard-coding height: 100vh without an override variable. Always expose --deck-height so chrome can sit above.

DeckHeader

3-column grid: wordmark · center nav (TOC · Scroll · Changelog with active-state highlighting + variant toggle pill when on a scroll deck) · right-aligned meta. Includes <GateScript /> so every page that uses the header inherits the gate.

• Reference: src/components/basics/DeckHeader.astro
• Common mistake: putting nav state into props instead of deriving from Astro.url.pathname. Active-state highlighting is path-driven so the chrome stays declarative across surfaces.

DeckNav

Fixed bottom-right counter + ‹ › nav. Presentational only — takes counter, prev, next, cycling, optional aria labels and an enableKeyboard flag. Used by SlideLayout (variant cycling) and the scroll-deck pages.

• Reference: src/components/basics/DeckNav.astro
• Common mistake: trying to make this component decide what to nav to. It’s a renderer; the page (or layout) computes prev/next from whatever registry it owns.

MetaTags

Comprehensive head metadata. Title (with suffix), description, robots (defaults to noindex), canonical URL (absolute), full OpenGraph block, Twitter card. Reads from lib/seo.ts registry.

• Reference: src/components/basics/MetaTags.astro
• Common mistake: forgetting to set site in astro.config.mjs. Without it, OG image URLs are relative and most preview unfurlers reject them.

GateScript

Inline <script is:inline> that redirects to / when the unlock flag is missing and the current path isn’t already the cover. Imported by DeckHeader so propagation is automatic.

• Reference: src/components/basics/GateScript.astro
• Common mistake: placing the redirect script in <body> rather than in <head> (or as the first child of <body>). It needs to run before any deck content paints.

Section components (teaser/, teaser-vN/)

One .astro file per slide per variant. Naming T{NN}-{Slug}.astro where NN is the slide number (zero-padded) and Slug is PascalCase.

• Reference: src/layouts/sections/teaser/T01-DisclaimerCover.astro through T17-FundTerms.astro
• Common mistake: putting content data inside section components. Shared data (team rosters, identity pillars, etc.) lives in src/data/ so all variants see the same content edits.

lib/ Registries

The pattern: every “set of things the chrome needs to know about” gets a dedicated registry module under src/lib/. The chrome reads from the registry; pages register themselves; new entries propagate everywhere automatically.

• lib/slides.ts — ordered slide registry (slug, number, title) + getSlideBySlug() + getNeighborHrefs().
• lib/scroll-decks.ts — scroll-deck variant registry (href, label, variantNumber) + isScrollDeckPath() + getScrollDeckCycle().
• lib/seo.ts — per-page metadata + helpers.
• lib/gate.ts — gate constants.
• lib/changelog.ts — markdown loader for context-v/changelogs/.

When you add a new surface, ask first: “should this be a registry?” If the answer is yes, the chrome will auto-discover entries and you’ll touch fewer files when the next entry shows up.

OpenGraph + SEO System

Detail beyond Step 8 above.

• Site URL (astro.config.mjs): site: process.env.SITE_URL ?? "https://{vercel-preview}.vercel.app". Set SITE_URL in Vercel’s env panel for production. Required for absolute canonical and OG URLs.
• Default OG image (/public/og_image__{Client}-Deck-{Fund}.png): one shared image referenced by DEFAULT_OG_IMAGE in lib/seo.ts. Provide exact pixel dimensions to the registry — preview platforms care.
• MetaTags component drops into every page’s <head> first. Two required props (title, description); everything else has sensible defaults.
• Per-page metadata lives in SLIDE_SEO, SCROLL_DECK_SEO, and STATIC_SEO registries inside lib/seo.ts. Pages look up their entry by slug or path.

Test by pasting a deck URL into Slack, iMessage, LinkedIn DM, X DM. The unfurl should show the brand image with the per-page title and description. If it shows a generic preview, the site URL is missing or the image path didn’t resolve to absolute.

Polite Gate Posture

Detail beyond Step 9 above.

The gate is politeness, not security. State this clearly when the client asks how the gate works, and offer the escalation path (Astro middleware + SSR + cookie verification) when the threat model warrants it. Capture the option matrix in context-v/explorations/Gate-Sensitive-Information-with-Simple-Code.md so future-you doesn’t have to redo the analysis.

The five-piece gate stops:

• Search engines and compliant crawlers (robots.txt + meta + header)
• Casual lurkers (“I clicked the link in Slack out of curiosity”)
• The client’s anxiety about what’s “just out there”

It does not stop:

• Anyone with a known URL using curl or a headless browser
• Determined scrapers with the URL list

For a brand-new Vercel preview URL with no inbound links, this is enough. Tighten when you ship to a real domain or share with a wider audience.

Changelog Hygiene

The frontmatter format that’s worked well across many entries:

---
title: "{Concise human title}"
lede: "{1-3 sentence summary that reads well at the top of /changelog}"
date_authored_initial_draft: YYYY-MM-DD
date_authored_current_draft: YYYY-MM-DD
date_first_published: YYYY-MM-DD
date_last_updated: YYYY-MM-DD
at_semantic_version: 0.X.Y.Z
status: Published
category: Changelog
tags:
  - Pattern-Names
  - With-Hyphens
authors:
  - Author Name
augmented_with: Claude Code (Opus 4.7, 1M context)
---

Body sections, in order:

• ## Why Care? — the stake. Why this change matters to the project, the audience, or the broader posture. This section is what /changelog extracts as the excerpt.
• ## What Was Built — the actual work. Subsection per pattern or per file group.
• ## Files Touched — a tree of changed paths with a one-line note per file. Keeps the diff legible without reading every commit.
• ## Open Items — known follow-ups, deferred decisions, things that surfaced during the work and are worth flagging.
• ## Reference — pointers to related changelogs, specs, explorations, live URLs.

The at_semantic_version field is the workspace’s narrative semver, independent of npm’s package.json#version. Bump minor for substantive feature additions, patch for polish.

package.json Metadata

Fill these standard fields generously — they’re free documentation that travels with the package and is read by IDEs, npm UIs, and future Claude sessions:

• description — one-paragraph what-it-is
• keywords — 10-20 tags for searchability
• license — UNLICENSED for private workspaces
• author (object: name, url) and contributors (array)
• homepage, repository, bugs
• engines (node, pnpm)

Plus a custom namespace block (e.g. "calmstorm": { ... }) with workspace-specific facts: fund / vehicle / domicile / audience, a siteSurfaces map of every URL pattern to its purpose, deployment notes, and pointers to canonical docs. npm/pnpm ignore unknown top-level keys; nothing breaks.

Reference: sites/calmstorm-decks/package.json.

Vercel Deploy Gotchas

• Standalone lockfile. After every pnpm --filter <site> add ..., cd into the site dir and run pnpm install --ignore-workspace --lockfile-only to regenerate the standalone pnpm-lock.yaml Vercel uses. Without this, Vercel builds with the workspace-root lockfile and errors out with missing dependencies.
• PUBLIC_DECK_CODE env var. Set in Vercel → Project Settings → Environment Variables. Without it, the gate falls back to the in-code default — works, but rotation requires a code change.
• SITE_URL env var. Set when the deck moves off the Vercel preview domain so canonical and OG URLs flip to the real domain on the next deploy.
• Vercel adapter output dir. .vercel/ is gitignored.

AI-Assisted Iteration Discipline

The patterns that proved out across the calmstorm-decks build week:

• Brief Claude with the foundation, not just the task. Before asking for a new variant or a new section, point Claude at theme.css, the narrative file for that slide, and an example existing section. The output quality is dramatically higher with the foundation in context.
• One coherent voice per scroll-deck variant. When inventing v2 or v3, decide on a single design language up front (e.g. “editorial print magazine”) and apply it consistently across all 17 sections. Mixed voices feel incoherent in scroll mode.
• Don’t peek at prior variants when inventing a new one. When asking Claude for variant N, instruct it to look only at the data files, narratives, and theme.css — not at variants 1..N-1. Otherwise the output drifts toward the existing variants instead of finding new territory.
• Registry-driven propagation. When you find yourself wiring the same surface into header + footer + nav + index page, stop and build a registry. The cost of the abstraction pays back on the second consumer.
• Changelog as you go. Write the changelog entry while the work is fresh, before committing. The “Why Care?” section forces you to name the stake, which often reveals when a change is bigger or smaller than you thought.
• Honest dispatches. When a pattern is theatrical (e.g. the polite gate), say so plainly in the changelog and exploration files. Future Claude sessions inherit the framing; clients trust documentation that doesn’t oversell.

Reference

• Reference implementation: sites/calmstorm-decks/ — every pattern in this spec is implemented there.
• Original spec: astro-knots/context-v/specs/Develop-a-Slides-only-Astro-Site-for-a-Fundraise-Process.md (the ask that became calmstorm-decks).
• Threat model for the gate: sites/calmstorm-decks/context-v/explorations/Gate-Sensitive-Information-with-Simple-Code.md.
• Audit trail for the build week: sites/calmstorm-decks/context-v/changelogs/ (April 30 → May 3, 2026).
• Parent guidance: astro-knots/CLAUDE.md — workspace structure, LFM consumption pattern, CSS token convention, design-system + brand-kit motion. Read first; this spec assumes that material.