More screenshots
Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+21 moreClaude CodeCursorCodex CLIMuseDots
Manus
Grok
Grok Bot+18 agentsUniversal Web Dev
Plan, build, debug and audit websites with your AI coding assistant. Get task-specific guidance, practical verification and a Launch Readiness Score with coverage, evidence and prioritized fixes. Free and open source.
Free
Universal Web Dev
Example session with this skill installed
Example prompt for the reference site shown below:
Use universal-web-dev to check the Saltrum reference site at http://localhost:8765 and report its Launch Readiness Score, coverage, blockers and top fixes. Report first, without changing code.
- Read your context and instructions
- Compiled the universal web dev
- Generated the UI component
Launch Readiness check: Saltrum reference site
Recorded example from the repository, dated 2026-10-27. This is the existing published report, not a new measurement of your site. Source: https://github.com/generosomm/universal-web-dev/blob/main/examples/reference-site/docs/readiness-check.md
A real check run with the Launch Readiness Score rubric. Nothing was changed before or after; the fixes below are left open on purpose so the example stays honest.
Launch Readiness Score: 70/100, Needs work Coverage: 21/25 (84%)
Mode: check Site: examples/reference-site, served locally Date: 2026-10-27
Conditions: Chromium (desktop app browser), desktop and 375×812 phone emulation, no network throttling
| Category | Score | Weight | Items checked | Status |
|---|---|---|---|---|
| Works as intended | 88 | 20 | 4 of 5 | Tide planner works, all 8 in-page links resolve, no console errors |
| Works for everyone | 79 | 25 | 7 of 8 | Every measured text passes contrast; one small tap target |
| Trust and safety | 75 | 25 | 2 of 3 | No secrets; third-party font and image requests without a privacy note |
| Speed | 67 | 15 | 3 of 4 | No blocking scripts, images sized and lazy; font CSS blocks rendering |
| Found and shared | 30 | 15 | 5 of 5 | Good title and description; no canonical, share image, structured data or sitemap |
Blockers: none.
Top fixes
| # | Finding | Evidence | Impact | Fix | Effort |
|---|---|---|---|---|---|
| 1 | No structured data | 0 JSON-LD blocks on / | Search engines and AI answers can't read the rooms, prices or address as facts | Add a LodgingBusiness JSON-LD block that matches the visible content | S |
| 2 | No sitemap.xml or robots.txt | Both return 404 | Crawlers get no index hints | Add a one-URL sitemap and a robots.txt that points to it | S |
| 3 | No canonical URL or share image | No link rel=canonical, no og:image | Shared links show no picture; duplicate URLs can split ranking | Add both to head once the domain is known | S |
| 4 | Third-party requests without a privacy note | Google Fonts and picsum.photos receive each visitor's IP address | May need disclosure for EU visitors | Self-host the font and the final photos, or add a short privacy note | M |
| 5 | Small tap target | "Ask about the whole house" link is 21 px tall at 375 px width | Harder to tap on a phone | Pad the link to at least 24 px, ideally 44 px | S |
Not run
| Item | Why | How to run it |
|---|---|---|
| HTTPS, redirects, mixed content | Not deployed | Re-run check against the deployed URL |
| Form failure path | The tide planner has no failure state worth forcing in a check | Covered by audit |
| Worst-case content (break-ui) | Not part of a quick check | Run audit |
| Core Web Vitals field data | No real visitors yet | PageSpeed Insights or CrUX after launch |
Not applicable
Server validation, spam limits and consent: the site has no server, sends no form data and stores nothing in the browser. Hreflang: one language.
Measurements behind the score
- Contrast: computed text and background colours for every visible text element; none below 4.5:1 (3:1 for large text).
- Keyboard: Tab reaches the skip link first with a 3 px solid outline; later links rely on the same focus style, partially checked.
- Phone (emulation only, so phone items count as partial at most): no sideways scroll at 375 px, date input text 20 px (no zoom on focus).
- Load (unthrottled, local): DOMContentLoaded 617 ms, load 2.1 s, mostly the remote placeholder image (1.7 s).
- 404: the local Python server returns a real 404 status but not the site's own 404.html; that depends on the host.
- The site's own tests: 11 passed (python -m unittest discover -s tests).
Suggested next steps
These steps are added for this listing, separate from the recorded report: address the prioritized findings, rerun the same check under comparable conditions, and check the deployed URL before launch. Use audit for broader coverage, compare for two real runs, or handoff for a client-facing report.
404.html
TSX · React component
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Plan, build, debug and check websites with a clearer process and evidence you can review.
Universal Web Dev is a free, open-source skill created by Merwin Generoso. It gives your coding assistant a small routing table that selects the guidance needed for each task: planning, design, implementation, debugging, testing, performance or search visibility.
What it helps you do
- Build landing pages, portfolios, dashboards and web-based business systems while respecting your chosen stack and existing design.
- Plan a new project before implementation, then work through complete features with acceptance criteria and test evidence.
- Diagnose bugs before editing and verify changes before calling them complete.
- Check phone layouts, keyboard access, contrast, links, forms, page speed and search metadata with the tools available to your assistant.
- Write client handoff reports and compare two real readiness checks.
Launch Readiness Score
Ask for a site check or audit to get a score from 0 to 100 across functionality, accessibility, trust and safety, speed, and search visibility. Reports include coverage, blockers, evidence and prioritized fixes. Checks that could not run are disclosed, and low-coverage scores are marked provisional. A score is a review aid, not a security or accessibility certification.
What this free package includes
The original Universal Web Dev wrapper, ten reference guides, a pinned dependency manifest, a Python companion installer, package checks, tests and community credits.
The ZIP does not bundle the 33 companion skills. The wrapper has a shorter built-in fallback when a companion is missing. For the full setup, ask your local tool-enabled assistant:
install or update universal-web-dev and all of its skills from:
https://github.com/generosomm/universal-web-dev
follow INSTALL.md, preserve my customized skills, and report what changed.
The installer downloads selected sources from reviewed, pinned commits. Its normal sync flow preserves customized and unmanaged copies. Community skills keep their own authorship and licenses, listed in CREDITS.md. No endorsement by those authors is implied.
Setup and scope
Claude Code and Codex have documented installer destinations. Other hosts need a verified skills directory and support for the required files and tools. Browser checks need a usable browser tool; optional Python Playwright is one supported route. The companion installer needs Python 3.10+.
Your AI provider, project runtime and hosting remain your own setup. The skill does not choose a framework, require a GitHub account, or include model/API credits. It grants no permission to deploy, spend money or change production data.
The original wrapper is MIT licensed. This is a community preview with a documented static-site trial; broad full-stack and cross-agent runtime compatibility has not been established.
Web projects only. The ZIP includes the wrapper, references and installer, not the 33 companion skills. Missing companions use a shorter fallback. Browser, device, provider and deployment checks depend on host tools and permissions; unrun checks are reported. Impeccable installs guidance-only, without its binary launcher. A readiness score is not certification. The documented end-to-end example is a static site, not proof of all full-stack workflows or all agents. AI services and hosting may have their own costs.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
This skill has been used 6 times.
Trust & safety
Security scan
Security scan passed 7 days ago
- Free with an Agensi account
Requires
GitHub
Python
Node.js
GitPermissions
Allowed Hosts
Read/write access is for the project the user selects. Shell access runs that project's commands and the optional companion installer/checkers. Browser/network access loads the site under review and downloads public pinned sources when installation is requested. No connected account is required. Environment access is limited to installer paths (CODEX_HOME/CLAUDE_CONFIG_DIR and home-directory resolution) and the maintainer update checker's optional GITHUB_TOKEN; it does not require users to supply or disclose credentials. Optional installation also writes to the chosen host skills directory and its sibling backup directory. Installation does not execute upstream skill code; later skill use may run documented helper scripts. Deployments, paid services, production data changes, commits and pushes need user authorization.