Build Awwwards-Quality Sites skill review: verdict, safety and alternatives

Build Awwwards-Quality Sites is Meng To's one-file skill for motion-led marketing sites with GSAP and one scroll engine. We rate it Recommended with caveats, 70/100, at Low confidence. Its honesty and accessibility rules are strong. It ships no code, and nobody has published a test of what it builds.

Get the skill: build-awwwards-quality-sites folder on GitHub · SKILL.md at the reviewed commit · skills.sh listing

What is the Build Awwwards-Quality Sites skill?

Build Awwwards-Quality Sites is Meng To's single-file skill for art-directing and building motion-led marketing sites. It covers GSAP choreography, one smooth-scroll engine, optional Three.js, honest assets, and accessibility checks before handoff.

An Agent Skill is a folder with a SKILL.md file that a coding agent loads on demand. Our guide on what Agent Skills are explains the format. This one lives in Meng To's MengTo/Skills library. A directory, agenticskills.io, names him as the author of Design+Code.

Awwwards is a web design award site. The skill borrows its name as a quality bar only.

The SKILL.md file is 7,216 bytes, about 1,804 tokens. It holds no code blocks. Every rule is an imperative bullet. A second file, agents/openai.yaml, sets a display name, a short description and a default prompt. The licence is MIT.

The file walks the agent through seven phases, in this order:

1. Set the art direction.
2. Build an honest asset system.
3. Compose the hero.
4. Build the motion system.
5. Add Three.js only with purpose.
6. Meet the quality bar.
7. Validate before handoff.

Four other "awwwards" skills share the name. They are different products from different publishers. We have evaluated none of them.

Numeric targets you see on other listings belong to those skills. The 60fps goal, the FCP under 1.5 seconds and the 8.5+ score threshold come from Awwwards Design. Meng To's file sets no such numbers.

Which rules does the skill enforce for motion, assets and accessibility?

The skill enforces exactly one smooth-scroll engine wired to ScrollTrigger. It requires final states under reduced motion and a complete static first frame. It bans fake testimonials, invented partners and award claims without proof.

GSAP is a JavaScript animation library. ScrollTrigger is its plugin that ties animation to scroll position. Lenis and Locomotive Scroll are smooth-scroll engines. Three.js is a WebGL library for 3D graphics in the browser. The skill names all of them and sets these rules:

• Run exactly one smooth-scroll engine, Lenis or Locomotive Scroll, wired to ScrollTrigger.
• Refresh ScrollTrigger after fonts and media load.
• Destroy timelines, triggers and the scroll engine on cleanup.
• Show final states when the visitor prefers reduced motion.
• Render a complete static first frame before any script runs.
• Keep an unsplit accessible name on every split headline.
• Provide poster images and handle a lost WebGL context.
• Ban fake testimonials, invented partnerships and generated people shown as real customers.
• Disclose any Logo Ipsum placeholder marks.
• Forbid award claims without proof.
• Ban SVG illustration drawn by the model itself.
• Run a production build before handoff.

Icons come from Iconify's Solar and Logos sets. The file gives no frame rates, timings or DPR caps.

Is Build Awwwards-Quality Sites worth installing?

Yes, with caveats, Build Awwwards-Quality Sites is worth installing for developers who know GSAP. It scores 70/100, Recommended with caveats, at Low confidence. Its honesty and accessibility rules are strong, but it ships no code snippets.

Read how we score skills for the weights.

Tessl is a registry that publishes quality reviews of Agent Skills. The skills.sh site is a directory that counts installs and lists security audits.

Thin evidence sets confidence to Low instead of lowering the score. The honesty and fallback rules are strong. The motion wiring rests on the model, because the file gives no setup snippet.

Compare it with other UI/UX and frontend design skills ranked on our design hub.

How many tokens does Build Awwwards-Quality Sites use?

Build Awwwards-Quality Sites loads 1,804 tokens per use, measured from file size. Our estimate for a typical task is 40,000 input and 15,000 output tokens. That task is one animated page with five to seven sections.

We assume two such tasks a week. Estimated hidden costs are image generation, real avatar photos, a second design skill and paid Aura.build.

What changed since our review?

The Build Awwwards-Quality Sites file has not changed since our 2026-09-30 review. The fourteen repository commits since then added other skills only. The verdict stays Recommended with caveats at 70/100.

The skill folder's only commit is still 6123969, dated 2026-07-25. The newer commits add game-combat VFX skills, 14 illustration skills and an HTML Demo skill. Stars rose from 6,495 to 6,862 and forks from 774 to 804, per the GitHub API on 2026-10-21. Issue #10 opened on 2026-10-02 with the title "skills" and an empty body. No release or tag exists.

Is Build Awwwards-Quality Sites safe to install?

Yes, Build Awwwards-Quality Sites is safe to install, with safety grade A. It is one Markdown file and a display manifest, with no scripts, network calls or credentials. Three skills.sh audits pass.

Our review of the files found nine facts:

• Files: SKILL.md is 7,216 bytes, with no references, scripts or assets.
• Manifest: agents/openai.yaml sets only a display name, short description and default prompt.
• Network: No network or environment-variable indicators appear.
• Code: The file has no code blocks at all.
• Named tools: GSAP, ScrollTrigger, Lenis, Locomotive Scroll, Three.js, Iconify, Logo Ipsum and Aura.build.
• Dependencies: The agent must pick and follow at least one other installed web-design skill.
• Writes: The agent writes only the site files it builds, then runs a production build.
• Audits: Gen Agent Trust Hub, Socket and Snyk all show Pass on skills.sh.
• Injection: The agent takes only high-level traits from reference sites and never copies assets or code.

That last rule is the only prompt-injection handling. Reference sites are still untrusted input. Our checklist on how to vet a skill covers what to check yourself.

What does the community report about Build Awwwards-Quality Sites?

The community has reported little about Build Awwwards-Quality Sites. Tessl rates its quality 70/100 and its content 73%. The skills.sh count was 2.0K installs at review, and no output test exists yet.

What are Build Awwwards-Quality Sites' known weaknesses?

Build Awwwards-Quality Sites has five known weaknesses. It gives no snippets or values for Lenis and ScrollTrigger wiring. It needs an image generator and a sibling skill, repeats rules, lacks an output test and leaves issues open.

1. Gives no snippets. Correct Lenis and ScrollTrigger wiring rests on the model's own knowledge. DPR caps and stagger timings are missing too, per Tessl's content score of 73%.
2. Needs extra tools. The agent needs an image generator or licensed PNG cutouts, because model-drawn SVG is banned. It needs real avatar photos and another web-design skill.
3. Repeats itself. The reduced-motion rule appears in sections 3, 4, 5 and 7.
4. Lacks an output test. Tessl lists no eval, and no issue or post reports a result.
5. Leaves issues open. Issue #2 and PRs #4, #6, #7 and #9 have no replies.

The worst failure is a broken site: two scroll engines fighting, or fake social proof shipped live. The rules target both, yet no test proves the agent follows them.

Which agents does Build Awwwards-Quality Sites work with?

No agent run of Build Awwwards-Quality Sites is verified. Meng To's repository claims that it works in Claude Code, Codex and Cursor. The folder ships a plain SKILL.md and an OpenAI agent manifest.

Nobody has reported a working run in any of the three.

How do you install Build Awwwards-Quality Sites?

You install Build Awwwards-Quality Sites with the skills CLI or by copying its folder into your skills directory. Three published commands exist. Then add a second web-design skill and image generation.

The README gives no install command, per SkillFoxx. Listings publish these forms:

1. Run the skills.sh form: npx skills add https://github.com/mengto/skills --skill build-awwwards-quality-sites.
2. Run the agenticskills.io form instead: npx skills add MengTo/Skills@build-awwwards-quality-sites.
3. Install the whole library with npx skills add mengto/skills when you want every Meng To skill.
4. Copy the folder agent-skills/web-design/build-awwwards-quality-sites into ~/.claude/skills/ as a manual option.
5. Add a second web-design skill, since the file asks the agent to follow one.

designmd.directory prints the skills.sh form with the repository name capitalised as MengTo/Skills. All forms point at the same folder. Our guide to set up skills in Claude Code covers folder paths and loading.

How do you build an Awwwards-style site with this skill?

To build an Awwwards-style site with this skill, ask for one premium landing page and name the skill. It plans the art direction and wires GSAP with one scroll engine. It then checks fallbacks before a production build.

1. Install a second web-design skill next to this one.
2. Enable image generation, or gather licensed photos and PNG cutouts.
3. Ask for one premium landing page and name the skill in the prompt.
4. Review the art direction the agent proposes before it writes code.
5. Let it wire GSAP, ScrollTrigger and exactly one scroll engine.
6. Test reduced motion and the WebGL fallbacks in a real browser.
7. Run the production build and check the static first frame.

Test it yourself before client work:

• Ask the same agent for one premium landing page with the skill.
• Ask again for the same page without it.
• Compare the Lighthouse results of both pages.
• Compare their reduced-motion behaviour.
• Count the scroll engines each page initialises.

What are the alternatives to Build Awwwards-Quality Sites?

The closest alternatives to Build Awwwards-Quality Sites are three evaluated skills. Taste Skill (62, Conditional) sets distinctive page direction. Emil Kowalski's Animate (77, Recommended) builds component motion, and Frontend Design (76, Recommended) suits calmer marketing pages.

Read the full Taste Skill verdict for its grade B caveats. Our review of Emil Kowalski's Animate skill covers component motion. The Frontend Design review suits teams who want calmer output. See the Landing Page Design verdict for why it scores 50.

Meng To's own cinematic-gsap-lenis-motion-system skill sits in the same library with 1.6K installs. We have not rated it.

Where is Build Awwwards-Quality Sites on GitHub?

Build Awwwards-Quality Sites lives on GitHub in the MengTo/Skills repository at agent-skills/web-design/build-awwwards-quality-sites, under the MIT licence. The skill file has one commit, dated 2026-07-25. The repository had 6,862 stars on 2026-10-21.

• Repository: MengTo/Skills, one contributor, MIT licence.
• Path: agent-skills/web-design/build-awwwards-quality-sites.
• Reviewed commit: 0c6169d09f8dff5dc26e57a8c33f027ffb3d28d9, evaluated 2026-09-30.
• File hash: sha256 1568e003d112e72df0ea2ec925829a5039b762c27174e063c192d7ccf7f4f9e6.
• Copies: 49 exact copies turned up in code search.
• Installs: 2.0K for this skill and 147.5K for the library on skills.sh.

Star counts differ by source. The repository had 6,495 stars at our review. agenticskills.io shows 6,835, and designmd.directory shows 4.8K. We found the repository's last commit dated 2026-10-01, a day after our review. The skill file itself did not change.

Common questions

Is Build Awwwards-Quality Sites free?

Yes, the skill is MIT-licensed. The optional Aura.build imagery service is paid, per SkillFoxx.

Does the skill include code?

No, it is written as directives only. It holds no snippets, timings or DPR values.

Can it build an Awwwards-style portfolio?

Portfolio sites with a Three.js shader hero are a claimed use case. No output test exists yet.

Is it for dashboards or app screens?

No, it targets marketing, editorial and portfolio sites. Product UIs and dashboards are a poor fit.