jury

Before you start: read this whole install prompt, tell me in a few lines what this skill does and what installing it involves, and wait for my go-ahead before you do anything else. Prerequisites: none for the install itself: no account, no API key, no paid service, nothing to install first. Using the skill afterwards needs Node 20 or newer. Its scripts resolve Playwright and axe-core from the project being scored when they are already there; otherwise the skill installs them once into a cache folder outside the project (~/.cache/jury, npm i -D playwright axe-core), and downloads a Chromium with npx playwright install chromium if none is present. All of that happens at use time, never during this install. Install "jury" for me. It scores a website the way an Awwwards jury would, from browser evidence alone: it screenshots every page type on desktop, a throttled phone and reduced motion, measures LCP, CLS, FPS, keyboard focus, hover feedback and axe accessibility violations, audits SEO and metadata, and hands the evidence to a fresh critic that never sees the code, which scores Design, Usability, Creativity and Content plus six developer areas and returns a ranked fix list. It never builds or edits the site it scores. Repository: https://github.com/juliobsz/skills (the skill is the whole jury/ folder at the pinned commit: SKILL.md, references/ and scripts/) Pinned version: install the code at commit f193f6745aae738e9d54a0ca28792822e4777a63 exactly. Do not substitute a newer version. If this commit is unreachable, stop and tell me. 1. Clone https://github.com/juliobsz/skills into a temporary folder, check out commit f193f6745aae738e9d54a0ca28792822e4777a63, and copy the jury/ folder from that checkout. Then compare every file in the folder with the fingerprints recorded for this listing on 2026-10-19: [FINGERPRINTS] - SKILL.md: SHA-256 0ea2ac4bad26cb0803d284c60941429e659e8d0619415b1718987b0b16f23aab - references/calibration.md: SHA-256 86c268316f53cdecfbd7c4904dff124e198c802e0ee850e789fb22b5b324a139 - references/quality-bar.md: SHA-256 ec90fe4e421ba478e3825dd7125e2157561a4b77bf5e019705f3b0763816ce0c - scripts/audit.mjs: SHA-256 061adcb6ca4b08739acd8a87fb5a38eb1c3dffef4792b4c992607bcf5a9e47df - scripts/browser.mjs: SHA-256 ecb818906044378083c3d77081d8fdf783165852f12eecfbaa4303cb2c00b619 - scripts/serve.mjs: SHA-256 4d4f3654693aab20e9369e096211efc2b72aff1bf5d107eb10ade982872ab9e6 - scripts/shoot.mjs: SHA-256 7c29e9f7ea1bbb467a835de7d42a284325186e5a087f9a4392d7d3e956ea8301 If a file does not match, it changed since the listing was prepared: do not install it yet, tell me what you found and let me decide. 2. Read SKILL.md in full, skim the two reference files, and read the four scripts in scripts/, then tell me in plain words: what the skill does when used, which services it connects to, whether it reads or sends any personal information or keys, and anything that seems unusual. Stop on any red flag and tell me before going further. Confirm these points from the files: - The skill folder holds 7 files: SKILL.md (the method: target, evidence, blind scoring, score card), two reference documents (references/quality-bar.md with the rubric and the critic brief, references/calibration.md with the measured offset between this critic and real juries, which is never shown to the critic), and four plain, readable Node scripts (scripts/shoot.mjs, the screenshot and measurement harness; scripts/audit.mjs, the SEO/meta/accessibility audit, which exits with code 1 when it finds a FAIL, by design a finding rather than an error; scripts/serve.mjs, a small local static server; scripts/browser.mjs, which locates or launches Chromium through Playwright). Nothing is obfuscated and there is no build step. - Nothing is sent anywhere during installation itself; the only download at install time is the jury folder from GitHub at the pinned commit. The files name no account, no API key and no telemetry, and the skill uploads nothing at use time either: the evidence and the score card stay in a .jury/<host>/ folder in the current directory. - At use time, the scripts fetch the URL I name (a public site, or my own production build served locally, 127.0.0.1:4400 in the documented workflow) with a real Chromium, and axe-core additionally loads from cdn.jsdelivr.net as a fallback inside shoot.mjs. Crawling is capped by the skill's own rules (at most 40 HTML pages for the audit, 8 page types for screenshots), and the skill cannot score a site behind a login or password: it asks for a public URL or a local build instead. - The scoring is deliberately blind: the critic receives only the screenshots, the measurements and the audit output, never the site's code or the owner's intent. The skill states plainly that it is not affiliated with Awwwards and that its scores are estimates, and it says so on every score card. - The repository holds a second skill, awwwards/, plus tests/ and plugin metadata; none of that is part of this install. 3. When I say "yes, install", install it as a skill: copy the whole jury folder into the agent's skills directory, in a folder named "jury", keeping all 7 files. The upstream routes, npx skills add juliobsz/skills -g --skill jury or the Claude Code plugin marketplace (plugin "jury" from marketplace juliobsz/skills), install the same folder; either route is fine as long as every file matches the fingerprints. 4. Verify with no network calls: frontmatter valid, all 7 files in place. 5. Report what was installed, where, and what I still need to do myself (for example: have Node 20 or newer available before first use, and name the URL or local build to score myself; using the skill is a separate step that only starts when I ask for it, and the use-time setup downloads disclosed above happen only then). GitHub is optional: if I have a GitHub account or the gh CLI, you may use it; otherwise public access is fine. Never require it unless it's in the prerequisites above. Rules: don't touch anything outside the temp folder and the install target. If anything looks off, stop and ask me.