{
  "markdown": "# Website Quality Checker\n\n**English** | [한국어](README_KO.md)\n\nA Claude Code skill that evaluates webpage quality based on [Google Search Quality Rater Guidelines (QRG)](https://static.googleusercontent.com/media/guidelines.raterhub.com/en//searchqualityevaluatorguidelines.pdf). Provide a URL or paste HTML/text, and get a comprehensive quality report with a grade (S/A/B/C/D) across 6 categories.\n\n## Key Features\n\n- **QRG-aligned evaluation** — Applies Google's official Search Quality Rater Guidelines systematically, not just SEO heuristics\n- **6-category inspection** — Page purpose, MC quality, website/author info, E-E-A-T, reputation, spam/abuse\n- **YMYL-aware** — Automatically detects YMYL topics (health, finance, legal, safety) and adjusts scoring weights accordingly\n- **Evidence-based scoring** — Every score must be backed by explicit evidence found in the content; inference is not accepted\n- **Anti-optimism bias** — Built-in cap rules, conservative defaults for unverifiable items, and a Devil's Advocate self-check to prevent inflated scores\n- **Actionable output** — Not just a grade, but a prioritized improvement plan with projected score gains\n- **Flexible input** — Accepts URLs (fetched via `web_fetch`) or directly pasted HTML/text\n- **Partial inspection** — Can run individual categories on demand (e.g., \"just check E-E-A-T\")\n\n## Overview\n\n### How It Works\n\nThe skill runs in 3 phases:\n\n**Phase 1: Discovery**\n\n1. Receive input (URL or pasted content)\n2. Fetch page via `web_fetch` if URL is provided\n3. Classify page type (informational, commercial, blog, forum, entertainment, government)\n4. Determine YMYL status (clear YMYL / borderline / non-YMYL)\n5. Lock in scoring weights based on YMYL determination\n\n**Phase 2: Analysis** 6. Run 6 categories sequentially (A → B → C → D → E → F), scoring each 0–100\n\n**Phase 3: Synthesis** 7. Compute weighted total score 8. Check for immediate D-grade triggers (harmful content, confirmed spam, fraud reputation) 9. Run Devil's Advocate self-check on all scores ≥ 75 10. Assign final grade and generate Markdown report\n\n### Inspection Categories\n\n| Category                      | Weight | What It Checks                                                       |\n| ----------------------------- | ------ | -------------------------------------------------------------------- |\n| A. Page Purpose & Harmfulness | 15%    | Is the purpose clear, beneficial, and non-deceptive?                 |\n| B. Main Content (MC) Quality  | 25%    | Effort, originality, skill, accuracy, appropriate depth              |\n| C. Website / Author Info      | 10%    | Operator transparency, author credentials, contact info              |\n| D. E-E-A-T                    | 25%    | Experience, Expertise, Authoritativeness, Trust                      |\n| E. Reputation                 | 10%    | Independent external sources, reviews, news coverage                 |\n| F. Spam / Abuse Detection     | 15%    | Copied content, scaled AI content, deceptive design, ad interference |\n\nWeights shift for YMYL topics — see [scoring-weights.md](skills/website-quality-checker/references/scoring-weights.md) for details.\n\n### Grading Scale\n\n| Score  | Grade | Google QRG Equivalent |\n| ------ | ----- | --------------------- |\n| 90–100 | S     | Highest               |\n| 75–89  | A     | High                  |\n| 55–74  | B     | Medium                |\n| 35–54  | C     | Low                   |\n| 0–34   | D     | Lowest                |\n\nCertain signals trigger an **immediate D grade** regardless of score:\n\n- Category A = 0 (harmful or deceptive purpose)\n- Category C = 0 + YMYL (no site/author info on YMYL topic)\n- Category E ≤ 10 (fraud/crime reputation)\n- Category F = 0 (confirmed spam)\n\n### Scoring Principles\n\n1. **Evidence-based** — Scores derive from what is explicitly present in the content, not inference. \"Seems like an expert\" does not count.\n2. **Unverifiable = conservative** — Missing info scores 35–40, not 50. \"Can't verify\" means \"can't trust.\"\n3. **Cap rules** — Any ❌ caps the category at 70; two or more cap it at 50.\n4. **Devil's Advocate self-check** — After scoring all 6 categories, every score ≥ 75 is re-examined. If evidence is insufficient, it drops to ≤ 60.\n\n## Plugin Structure\n\n```\nwebsite-quality-checker/\n├── .claude-plugin/\n│   └── plugin.json                          # Plugin manifest\n├── skills/\n│   └── website-quality-checker/\n│       ├── SKILL.md                         # Skill definition (3-phase workflow)\n│       └── references/                      # Loaded on demand during inspection\n│           ├── google-qrg-summary.md        # QRG core: YMYL, MC quality, reputation\n│           ├── eeat-criteria.md             # E-E-A-T evaluation questions\n│           ├── lowest-quality-signals.md    # Immediate Lowest triggers\n│           └── scoring-weights.md           # Weight tables & score formula\n├── README.md                                # English\n├── README_KO.md                             # Korean\n├── CHANGELOG.md\n└── LICENSE\n```\n\n**Progressive disclosure**: The skill body (`SKILL.md`) stays under 500 lines. Detailed criteria are in `references/` and loaded only when the relevant category is being inspected, keeping context usage efficient.\n\n## Usage\n\nAsk Claude Code naturally:\n\n```\n\"Check the quality of https://example.com\"\n\"Is this site trustworthy?\"\n\"Run a page quality analysis on this URL\"\n\"Evaluate E-E-A-T for this page\"\n\"Is this a spam site?\"\n\"Just check MC quality and spam signals for this page\"\n```\n\nYou can also paste HTML or text directly:\n\n```\n\"Evaluate the quality of this content: [paste text]\"\n```\n\nWhen no URL is provided, the reputation category (E) is marked as \"unverifiable\" and scored conservatively (35 points).\n\n### Partial Inspection\n\nRequest specific categories only:\n\n- \"Just check E-E-A-T\" → runs Category D only\n- \"Is this spam?\" → runs Category F only\n- \"Can I trust this site?\" → runs Categories C + D + E\n\nPartial inspections include only the requested categories in the report, with the overall score labeled as \"partial.\"\n\n## Output\n\nGenerates a Markdown report containing:\n\n### Report Sections\n\n1. **Header** — Target URL, inspection date, page type, YMYL status\n2. **Overall Grade** — Grade (S/A/B/C/D), score out of 100, visual progress bar\n3. **Score Summary Table** — All 6 categories with weights, scores, weighted contributions, and status labels\n4. **Category Details** — Per-category breakdown with ✅/⚠️/❌ verdicts and evidence citations for each item\n5. **Improvement Action Plan** — Prioritized into 3 tiers:\n   - Immediate fixes (high impact, low effort)\n   - Mid-term improvements (high impact, medium effort)\n   - Strategic improvements (long-term)\n6. **Projected Improvement** — Before/after score estimates per category if all recommendations are applied\n7. **Devil's Advocate Log** — Documents any score adjustments made during self-check\n8. **Key Takeaway** — 1 strength + 1–2 weaknesses in ≤ 3 sentences\n\n### Sample Grade Output\n\n```\n## Overall Grade: B (68/100)\n\n0         20        40        60        80        100\n|---------|---------|---------|---------|---------|\n██████████████████████████████████░░░░░░░░░░░░░░░░░ 68/100\n                                  ^\n```\n\n## Installation\n\n```bash\ngit clone https://github.com/llqqssttyy/website-quality-checker.git\n```\n\n## Reference Documents\n\nThe skill loads these on demand during inspection:\n\n- [google-qrg-summary.md](skills/website-quality-checker/references/google-qrg-summary.md) — YMYL classification, MC quality axes, reputation standards\n- [eeat-criteria.md](skills/website-quality-checker/references/eeat-criteria.md) — Detailed E-E-A-T evaluation questions per dimension\n- [lowest-quality-signals.md](skills/website-quality-checker/references/lowest-quality-signals.md) — Immediate Lowest-quality triggers (harmful, untrustworthy, spammy)\n- [scoring-weights.md](skills/website-quality-checker/references/scoring-weights.md) — Weight tables, YMYL adjustments, score calculation formula\n\n## License\n\n[MIT](LICENSE)\n",
  "bytes": 8040,
  "sha": "1baec9756b9aab20ccce8775e9e1f86f50114dc26552bcf8a16ac2a501227cec",
  "repo_slug": "llqqssttyy/website-quality-checker",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_llqqssttyy_website_quality_checker_websi_6ac5243a/readme"
}