long-form
Write datasheets and whitepapers. Use when the user asks for a datasheet, whitepaper, solution brief, technical paper, or other long-form asset that isn't a blog post or case study. Also trigger on 'w
SKILL.md.claude/skills/long-form/SKILL.md
GitHubLong-Form Content
Datasheets and whitepapers: the assets a buyer downloads, forwards, or prints. This skill covers the writing. Design and export are handed to brand-design and html-to-pdf.
Before Writing
- Confirm the facts you need exist in
/brain/truth.md. If they don't, list the gaps and ask before drafting. - Confirm the target reader matches a persona in
/brain/personas.mdor the ICP in/brain/positioning-and-messaging.md. - Read "Words We Use" / "Words We Avoid" in
/brain/positioning-and-messaging.mdand use the brain's exact category and product labels. - Read
/brain/voice-and-tone.mdif it exists.
Datasheets
- Structure: Problem > Solution overview > Key capabilities (3-5 bullets) > How it works > Technical specs > Proof points > CTA
- Length: One page, two-sided at most
- Tone: Concise and scannable. Bullets over paragraphs.
- Lead with the customer problem, not the feature list.
- Accuracy: every spec and capability comes from
truth.md. Mark anything else[VERIFY].
Whitepapers
- Structure: Executive summary > Problem definition > Market context > Solution approach > Technical deep dive > Results/proof > Conclusion
- Length: 2,000-4,000 words
- Tone: Authoritative and educational. Teach, don't sell.
- Data: Use industry data, benchmarks, or research with sources. Mark unverified stats
[VERIFY]. - Visuals: Add
[DIAGRAM: description]placeholders where a figure would carry the argument.
Production pipeline
- Write the content (this skill). Draft in Markdown under
marketing/long-form/[slug]/. Get the user's approval on the content before design. - Build as HTML/CSS (
brand-design). Produce a multi-page branded HTML document from the approved content. Typography, layout, colors, and diagrams are rendered in HTML/CSS. - Export (
html-to-pdf). Export the HTML to a print-quality vector PDF. Optionally push to Figma for team edits.
The HTML version is the source of truth for content and layout. Figma is for post-production polish, not authoring.
Audience Calibration
Technical readers (engineers, platform teams): precise technical language, architecture, code or configuration snippets where they help, integration points and operational impact. No superlatives.
Business readers (VPs, directors, C-suite): lead with outcomes (cost, time, risk, team velocity), translate capabilities into business value, prefer proof points and metrics over mechanism, keep it short.
SEO (web-published assets only)
If the asset will live on a web page, not only as a PDF: primary keyword in the title, first paragraph, and at least one H2; a 150-160 character meta description; a short descriptive slug. For deeper work, load seo-geo.
Review Checklist
Before delivering a draft:
- Every product claim is sourced from
/brain/truth.mdand cited inline (*(source: truth.md)*) - Messaging aligns with
/brain/positioning-and-messaging.md - No invented features, metrics, customer names, or quotes; unapproved quotes are
[APPROVED QUOTE NEEDED] - Unverified claims are marked
[VERIFY] - Headings are clear and descriptive, not clever
- One specific CTA
- Run the
copy-editingskill, then Vale
Related Skills
- brand-design: Phase 2 HTML/CSS build
- html-to-pdf: Phase 3 PDF export
- product-marketing: One-pagers, comparison documents, battlecards, objection handling
- copy-editing: Editing pass before delivery
- case-studies: Customer stories