humanmade/accelerate-ai-toolkit/accelerate-design
Claude Code and agent skill by humanmade/accelerate-ai-toolkit
- community
Checking that the SKILL.md file is still in the repository.
Design a new on-brand version of a block or a fresh section — "make me a variant of the hero", "design a new pricing section", "compose an on-brand version", "give me a bolder hero". Produces ready-to-use content; offers to put it in a test but never forces one. NOT for running or tracking an A/B test (use accelerate-test) or a multi-round optimization loop (use accelerate-evolve).
Install
npx vskill@latest install humanmade/accelerate-ai-toolkit/accelerate-designSKILL.md
Accelerate — Design an on-brand variant or section
You author a brand-fitting version of a block, or a brand-new section, on demand. This is test-optional: you produce the content and offer to set up an A/B test, but creating a test is a separate, confirmed step.
Route elsewhere when: the user wants to set up / monitor / end one A/B test → accelerate-test. The user wants to improve a block over multiple rounds, automatically → accelerate-evolve. This skill is the single-shot "make me a version" primitive both of those build on.
Flow
- Load the brand pack. Follow
docs/brand-pack.md— the whole-site grammar: global styles + the usage-grounded structure library (real section fragments harvested from published pages, with synced patterns confirmed by page usage) + media. Never one page. Reuse the cached pack if it's fresh. - Take the brief. Which block or section, and what the user wants it to do (or "a new <kind> section"). If targeting an existing block, read its current content first:
accelerate/get-variantsif it already has variants; otherwise WP-CLI (wp post get <id> --field=content) or ask the user to paste it. Do not reconstruct it from memory. - Compose by recombining the site's real fragments. Work from the brand pack's
fragments[]— the actual block markup the site uses, not summaries — and recombine and extend them; don't re-skin the control. Prefer high-confidence (in-use) fragments; treat low-confidence (registered-but-unused) ones with suspicion. Keep the fragments' semantic classes and preset-slug tokens intact as you splice. Aim for a composed redesign (Visual Score 2); reach for a ground-up reimagining (Visual Score 3) on a bold ask. Build it as a block tree, then serialize. Reference design tokens by preset slug only; reuse real media from the site; never invent facts, numbers, or attachment IDs. - Run the Block Runner pre-flight, then validate silently against `docs/design-standards.md`. Before showing a serialized Gutenberg candidate, follow
docs/block-runner.md:validate; on exit1,fixthen re-validate; use repaired markup only after the second check passes. Rework or withhold a candidate that remains invalid — never show it or hand it to a write ability. Ifnpxis unavailable, times out, or headless Gutenberg cannot boot, use the §2 manual checks and tell the user automated markup validation was unavailable; exit2is a prominent broken-invocation warning before the same fallback. Then check slug-first (§1), anti-pattern bans (§3), differentiation rubric ≥3/7 with no zeros (§4), and no AI-slop copy (§5). For full-bleed sections / whole-page work, also honour the rendering & editability contract (§15: forgiving primitives overwp:cover, editor styles, full-width alignment, content width). Rework anything that fails before showing it. - Present it, then offer the test. Show what it is, the concept it argues, and the real fact it's grounded in. Then offer the next step — set it up as an A/B test (hand to
accelerate-test), or just hand back the content to use as-is. Never create a test without explicit confirmation.
Presenting
Lead with the concept, not the markup. Example:
Here's a bolder take on your hero — an outcome-led version.Your current hero leads with what the product is. 62% of these visitors arrive from searches about slow sites, so this version argues the outcome instead — "Fix your slow site in one afternoon" — staged as a full-bleed section with one high-contrast call-to-action, using your own heading font and brand colours. (Grounded in: 62% of hero traffic arrives from speed-related searches.)Want me to set this up as an A/B test against your current hero, or just hand you the content to drop in?
Show the variant's copy in the presentation. Keep the structure description in plain English — never paste raw block markup at the user or name a capability. Apply the router's full terminology ban to the entire answer, including setup and verification notes; never call a site operation an "ability".
Rules
- Whole-site grammar, never one page. Always load the brand pack (
docs/brand-pack.md) before composing. - Compose, don't re-skin. Recombine the site's own sections; a real version is a different experience, not the control with new words. Apply the bold-by-default doctrine in
docs/design-standards.md§9. - Slug-first, real media, no invented facts. Every token is a preset slug; every image already exists on the site; placeholder copy where you'd otherwise guess a number.
- Test-optional. Produce and offer. Never call create-ab-test (a mutation) without explicit confirmation — hand to
accelerate-testonce the user opts in. - No AI-slop copy (
design-standards.md§5) and no developer jargon in anything the user reads.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/accelerate-design/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/humanmade/accelerate-ai-toolkit/accelerate-design)