AI Agents Get an SEO Department in a Repo

What it does

This repo is a library of twenty agent skills—zero-dependency Markdown files plus a thin Bash runner—that teach Claude Code, Cursor, Codex, and compatible agents how to perform SEO and Generative Engine Optimization (GEO). Skills span the full lifecycle: research, content building, on-page and technical optimization, and monitoring. Each skill follows a rigid activation contract (Quick Start, Skill Contract, Handoff Summary, Data Sources, Instructions, Reference Materials, Next Best Skill) so agents don’t hallucinate their way through an audit.

The interesting bit

The author treats prompt engineering like software architecture. Four cross-cutting “protocol” skills—content-quality-auditor, domain-authority-auditor, entity-optimizer, and memory-management—provide a governance layer. The memory-management skill is particularly fussy: it maintains HOT, WARM, and COLD project memory, compiles wiki pages from raw WARM files, surfaces contradictions with (a)/(b)/(s)/(i) prompts, and lets you retire or recover old context. It’s basically a CMS for an agent’s short-term memory.

Key highlights

• 20 skills covering keyword research, competitor analysis, SERP analysis, content-gap analysis, SEO/GEO writing, schema markup, technical audits, rank tracking, and alerting.
• Content quality is gated by an 80-item CORE-EEAT benchmark; domain trust is gated by a 40-item CITE benchmark.
• Every skill works at Tier 1 with user-provided data, so you aren’t locked into proprietary connectors.
• A tiered memory system (HOT/WARM/COLD) with a wiki compilation layer manages long-running project context.
• Packaged for 35+ AI agent hosts including Claude Code, Gemini CLI, Qwen Code, and Kimi Code CLI.

Caveats

• This is fundamentally a structured prompt library and methodology, not a traditional code package; engineers hunting for heavy automation scripts will find mostly Markdown contracts and a small Bash runner.
• The README explicitly disclaims that these skills do not guarantee rankings, AI citations, traffic, legal compliance, or business outcomes.
• Compatibility claims for 35+ hosts are documented, but support levels and manifest ownership live in separate marketplace files, so your mileage may vary by platform.

Verdict Content strategists and SEO specialists who already live inside Claude Code or Cursor will get the most value; developers looking for a headless API or self-hosted ranking engine should look elsewhere.