Building a programmatic SEO site with an AI coding assistant sounds straightforward. Tell it what you want, let it generate the pages. But anyone who has tried this knows the reality: the assistant doesn't understand crawl budgets, internal linking patterns, thin content thresholds, or Google's scaled content abuse policies. You end up with thousands of pages that look fine on the surface but fail in production.
pSEO Skills is an open source project that solves this. It's a collection of 12 specialized skill files that you drop into your AI coding assistant's configuration directory. Each skill teaches the assistant a specific aspect of programmatic SEO, from data architecture to quality gates to 100K-page scaling.
What's in the toolkit
The project includes 12 skills, each focused on a different part of the pSEO pipeline:
| Skill | What it does |
|---|---|
| pseo-discovery | Scans your codebase for pSEO opportunities |
| pseo-audit | Assesses your site's readiness for scaling (read-only) |
| pseo-data | Designs structured data architecture with a two-tier system |
| pseo-templates | Creates unique, intent-matched page templates |
| pseo-metadata | Implements dynamic meta tags, titles, and descriptions |
| pseo-schema | Adds JSON-LD structured data per page type |
| pseo-linking | Builds hub-and-spoke internal linking strategies |
| pseo-performance | Optimizes Core Web Vitals at scale |
| pseo-llm-visibility | Enhances content for AI search citation |
| pseo-quality-guard | Detects thin content, duplicates, and abuse patterns |
| pseo-scale | Handles architecture for 10K to 100K pages |
| pseo-orchestrate | Coordinates all skills in the correct execution order |
Each skill is a standalone Markdown file. No dependencies, no build step, no lock-in.
How it works
The skills are written as structured instructions that AI coding assistants can follow. You place them in the appropriate directory for your tool:
- Claude Code:
.claude/skills/ - Cursor:
.cursor/rules/(rename to.mdc) - GitHub Copilot:
.github/copilot-instructions.md - Windsurf:
.windsurf/rules/
Once installed, you can invoke skills directly. Ask your assistant to "run pseo-orchestrate with full pipeline" and it will execute each skill in dependency order, or invoke individual skills for specific tasks like "run pseo-quality-guard on the city pages."
Why this matters
Most AI coding assistants have general programming knowledge but no specialized pSEO expertise. They don't know that:
- Publishing 5,000 pages at once triggers spam reviews
- Pages sharing more than 60% content get flagged as thin
- MinHash/LSH algorithms are the efficient way to detect near-duplicates at scale
- Graduated rollout (100 pages, then 1,000, then 10,000) is safer than batch publishing
- AI crawler configuration and llms.txt affect how LLMs cite your content
The skills encode this knowledge so the assistant makes the right decisions automatically.
Quality guardrails
The pseo-quality-guard skill is particularly interesting. It uses MinHash and SimHash algorithms to detect content similarity across your pages without the O(n^2) cost of comparing every page pair. It also checks for:
- Data sufficiency gating (prevents generating pages when there isn't enough unique data)
- YMYL risk assessment for sensitive topics
- E-E-A-T signal verification
- Scaled content abuse pattern detection
This means your AI assistant won't just generate pages. It will refuse to generate pages that would hurt your site.
The graduated rollout system
The pseo-scale skill implements a phased publication strategy:
- Pilot batch (100 pages): Highest-quality pages, monitor for 2-4 weeks
- Growth phase (1,000 pages): Expand if indexation and ranking signals are healthy
- Scale phase (10,000 pages): Systematic expansion with crawl budget management
- Full scale (100,000 pages): Architecture for sites that need massive coverage
Each phase has specific quality gates that must pass before the next phase begins. The assistant tracks crawl budget, indexation rates, and ranking signals to decide when to proceed.
Framework support
The primary target is Next.js App Router, but the skills are designed to adapt to other frameworks:
- Astro: Static-first approach, great for content-heavy pSEO
- Nuxt: Vue-based alternative with similar SSG/SSR capabilities
- Remix: Nested routing aligns well with pSEO hub-and-spoke patterns
- SvelteKit: Lightweight option for performance-sensitive deployments
The data layer and template patterns translate across frameworks. Only the routing and rendering details change.
Getting started
- Clone or download the skills from GitHub
- Copy the skill files to your assistant's configuration directory
- Start with
pseo-discoveryto identify opportunities in your existing codebase - Run
pseo-auditto assess readiness - Use
pseo-orchestrateto execute the full pipeline
The project is MIT licensed and actively maintained. Whether you're building your first pSEO site or scaling an existing one, these skills give your AI coding assistant the expertise to do it safely.
