RankWise AI Agent

RankWise - Professional SEO Content Engine

[ English | Русский ]

Rank higher by writing for humans first, search engines second.

An AI skill system that generates, rewrites, and audits content against 49 SEO ranking factors - covering everything from keyword placement to schema markup, E-E-A-T signals, and featured snippet optimization. Skill class: orchestrator - self-contained SKILL.md plus on-demand references.

---

Scope

RankWise is a content-level SEO engine - it optimizes text, meta tags, alt texts, URLs, and content structure. It does not measure page speed, Core Web Vitals, server configuration, SSL, or cookie compliance. For those, use Lighthouse, PageSpeed Insights, or Screaming Frog.

---

What is this?

RankWise is a system prompt (skill) for any LLM - GPT, Claude, Gemini, DeepSeek, or any capable model. It transforms the AI into a world-class SEO strategist and content engineer who:

• Writes content that scores 46+/49 on the ranking factors checklist
• Rewrites existing content to fix SEO gaps (from URL or pasted text)
• Audits content against 49 measurable ranking signals
• Generates optimized meta tags, image alt texts, and URL slugs
• Creates complete SEO content briefs and editorial plans

It covers every critical SEO signal: readability scoring, schema markup, E-E-A-T signals, featured snippet targeting, topic clustering, and LSI keyword strategy.

---

How it works

1. Load SKILL.md as a system prompt into any LLM.
2. Give a task: "SEO article. Keyword: email marketing." / "Write about email strategy + meta. Keyword: email." / "Audit this content." / "Full package for my landing page."
3. Get output - only what you asked for. No extra modules. No fluff.

Six modes with modular output:

• Generate - Create SEO-optimized content from scratch (keyword → article, topic + keyword → article + meta)
• Rewrite - Transform existing content (URL or text) with full SEO optimization + AI-marker removal
• Audit - Score any content against 49 factors with weighted scoring, grade, prioritized fixes
• Meta-Only - Generate/optimize meta titles, descriptions, OG tags, alt texts, URL slugs
• Brief - Create a complete content brief - keywords, structure, linking plan, media plan
• Quick-Fix - Fix individual factors: «fix my title length», «add keyword to H2», «reduce passive voice»

Context-aware defaults:

• Keyword only → article body (no meta, no alts)
• Topic + keyword → article + meta
• Topic + keyword + images → full package (article + meta + alts)
• Any combination can be explicitly requested: «just the article», «meta only», «full package»

---

The 49-Factor SEO Scorecard

Every piece of content is scored against 49 factors across 5 categories with weighted multipliers (CRITICAL=×3, HIGH=×2, MEDIUM=×1, LOW=×0.5). Below are the highlights - full checklist in shared/checklist.md.

Category	Factors	Highlights
Keyword Placement (K1-K12)	12	Focus keyword, title, meta, URL, headings, density, cannibalization
Content Quality (C1-C14)	14	Word count, readability, heading hierarchy, passive voice, paragraph structure
Linking (L1-L9)	9	Internal links, external links, DoFollow, anchor variety, orphan prevention
URL & Technical (T1-T8)	8	URL length/structure, title/description length, schema, canonical, OG tags
Advanced Signals (A1-A6)	6	Featured snippet, E-E-A-T, LSI keywords, freshness, mobile readability, breadcrumbs

Weighted scoring: CRITICAL factors carry ×3 weight (any CRITICAL failure caps maximum grade at B). N/A factors excluded from denominator. Full methodology in shared/checklist.md.

---

Architecture

RankWise/
├── skills/rankwise/                ← runtime skill package
│   ├── SKILL.md                    ← Main orchestrator (6 modes, 49 factors)
│   ├── shared/                     ← Checklist, keyword rules, power words, schema
│   ├── scenarios/                  ← 11 procedural playbooks
│   ├── examples/                   ← Annotated examples
│   ├── validator/                  ← Python metrics + fix orchestrator
│   └── agents/openai.yaml          ← OpenAI agent manifest
├── tests/                          ← Validator unit tests (dev)
├── benchmarks/                     ← 5 reference texts (dev)
├── install.py                      ← Installer for Claude, OpenCode, Codex, .agents
├── README.md
├── CHANGELOG.md
└── LICENSE

---

Validator (Python)

The skills/rankwise/validator/ directory contains a Python tool that computes measurable SEO metrics from text - complementing the LLM-based scoring with deterministic calculations:

pip install -e .

# Full report
rankwise-validator --text "Your content..." --keyword "seo" --title "My Title" --meta "Description"

# Read from file
rankwise-validator --file article.txt --keyword "seo strategy" --title "SEO Guide"

# Checklist scores only (machine-readable JSON)
rankwise-validator --file article.txt --keyword "seo" --title "Title" --checklist --json

What it computes: word count (C1), paragraph metrics (C2), readability - Flesch-Kincaid, Gunning Fog, SMOG, ARI, Coleman-Liau for EN only (C8), passive voice ratio per language (C9), transition word ratio per language (C10), sentence length variety (C11), consecutive sentence starts (C12), keyword density (K10), keyword placement (K2, K6, K7), title/meta length with per-language ranges (T3, T4), power words in all 9 languages (C6).

Multi-language support: --lang en|ru|uk|de|fr|es|pt|it|pl. Passive voice, transition words, and power words use language-specific patterns for all 9 languages. Readability (C8) is computed deterministically via ReadSightPy (TeX/Liang syllable counting) for all 9 languages - EN uses Flesch-Kincaid Grade, DE Wiener Sachtextformel, ES Fernández-Huerta, IT Gulpease, FR/RU/PT Flesch Reading Ease, UK LIX, PL FOG-PL. Title/meta length targets adjust per language.

---

Benchmarks

benchmarks/ contains 5 reference texts with known SEO properties for regression testing across versions:

#	File	Expected Result
1	01-well-optimized.txt	Mixed - keyword in body, title mismatch (plural)
2	02-keyword-stuffed.txt	D-F: 23% density, word count fail
3	03-too-short.txt	D: 69 words, missing meta
4	04-no-meta.txt	D: no title/meta, short content
5	05-product-page.txt	C+: best density & keyword placement

# Run a single benchmark (omit --title/--meta if not applicable)
rankwise-validator --file benchmarks/01-well-optimized.txt --keyword "content marketing strategy" --title "7 Proven Content Marketing Strategies That Actually Work in 2026" --meta "Content marketing strategy guide: 7 proven tactics." --checklist

Coverage: The validator computes 16 of 49 factors (K1-K2, K6-K7, K10, C1-C2, C5-C6, C8-C12, T3-T4). Remaining 33 factors require LLM judgment or web access. See benchmarks/README.md for details.

---

Quick Start

Keyword only → article body:

"SEO article. Keyword: email marketing strategy."

Topic + keyword → article + meta:

"Write about email marketing strategy for SaaS founders. Keyword: email strategy. Language: en."

Full package (article + meta + alts):

"Full package about content marketing. Keyword: content strategy. Images: dashboard, chart, team photo."

Meta tags only:

"Generate meta title and description for my SaaS landing page. Keyword: project management software."

Image alt texts only:

"Generate SEO alt texts for these 5 images: [array]. Focus keyword: email marketing."

Full SEO audit:

"Audit this content for SEO: [paste text]. Focus keyword: project management software."

Content brief:

"Create an SEO content brief for a pillar page. Target keyword: content marketing strategy."

Rewrite from URL:

"Rewrite this article for SEO: [URL]. Focus keyword: SEO checklist."

Joint with human-ai:

"SEO-rewrite this article using RankWise, then humanize it with human-ai. EN. Keyword: seo audit."

Joint with MindFluence:

"RankWise SEO brief for SaaS landing page, then MindFluence from that brief. Expert-calm tone."

---

Integration with human-ai and MindFluence

RankWise is designed to work with, not against, these skills:

• human-ai - Text Humanization Engine. RankWise handles SEO structure → human-ai humanizes the voice. Monitor keyword density and heading keywords after humanization.
• MindFluence - Cognitive Bias Marketing Engine. RankWise provides SEO structure → MindFluence applies cognitive bias persuasion. Monitor for keyword over-stuffing in bold-sell tone.
• Triple pipeline: RankWise (structure) → MindFluence (persuasion) → human-ai (voice) → RankWise Audit (final check).

See SKILL.md → Integration with Other Skills for the full compatibility guide.

---

Requirements

• Any capable LLM with a system prompt / custom instructions field
• For full skill functionality: a skill system that loads files from a folder (OpenCode, Claude Code)
• SKILL.md is self-sufficient - contains all core rules for standalone use. skills/rankwise/shared/ and skills/rankwise/scenarios/ add professional depth (scoring formulas, AI-marker lists, per-language readability, schema templates, content-type blueprints)
• No API keys, no tools, no dependencies - pure prompt engineering
• Works in any language; per-language parameters for 9 languages (EN, RU, UK, DE, FR, ES, PT, IT, PL)
• Optional: Python 3.10+ + readsight for deterministic metric validation via skills/rankwise/validator/. Without readsight, C8 is N/A and the other 15 factors still work.

---

Installation

The installer copies skills/rankwise/ into the selected agent directory:

python install.py --target all
python install.py --target all --with-deps   # also installs readsight
python install.py --check-deps               # verifies the validator dependencies
python install.py --list
python install.py --check
python install.py --target project --dry-run

Without readsight the validator still runs: C8 readability returns N/A, and the other 15 factors are computed.

OpenCode

rankwise/ → .opencode/skills/          (project)
          → ~/.config/opencode/skills/  (global)

Claude Code

rankwise/ → ~/.claude/skills/

Any LLM (standalone)

Copy contents of skills/rankwise/SKILL.md as a system prompt. Add shared/ and scenarios/ files for richer detail.

---

License

MIT | Copyright Yevhen Leonidov

---

GitHub Description

SEO Content Engine for LLMs - generate, rewrite, and audit content against 49 ranking factors with multilingual support.

GitHub Tags

ai ai-agents content content-marketing content-strategy content-strategy-optimization llm llm-agent multilingual prompt-engineering seo seo-audit seo-optimization seo-tools

Skills.sh: https://www.skills.sh/madeval/rankwise