geo-skill: A Generative Engine Optimization Toolkit for the AI Search Era

geo-skill is an open-source GEO (Generative Engine Optimization) skill kit developed by the vespid-ai team. It provides reusable GEO skills and practical CLI tools for development teams, supporting direct use by agents such as Hermes Agent, Claude Code, and Codex. This toolkit aims to solve the problem of vague and hard-to-implement GEO suggestions by transforming abstract concepts into actionable checklists and automated tools, helping products improve their visibility in AI search tools like ChatGPT Search and Doubao.

With the rise of generative AI search tools like ChatGPT Search and Doubao, traditional SEO is undergoing changes. Users no longer rely on keyword matching to obtain information; instead, they get comprehensive answers through AI queries, spawning the GEO field. GEO focuses on 'AI referability'—whether content is structured, authoritative, and timely, making AI willing to cite it when answering related questions, which differentiates it from SEO's ranking orientation.

geo-skill is an open-source GEO skill kit and CLI toolset, with core features including multi-agent compatibility (the same skill supports Hermes, Claude Code, and Codex). Its design philosophy focuses on 'actionable GEO', converting vague optimization suggestions into specific checklists and automated tools to solve the problem of difficult GEO implementation.

geo-skill provides 16 GEO skills covering the entire product lifecycle:

1. Search engine adaptation layer: openai-chatgpt-search, doubao-bytespider, geo-bing-webmaster-foundation
2. Website architecture and content layer: geo-site-readiness, geo-structured-data-software-sites, geo-content-modeling, geo-homepage-positioning, geo-feature-pages
3. User trust and conversion layer: geo-pricing-pages, geo-faq-coverage, geo-docs-help-center, geo-changelog-freshness, geo-comparison-pages, geo-trust-and-entity-pages
4. Open-source and distribution layer: geo-oss-repo-geo, geo-launch-distribution

geo-skill provides a zero-dependency Python CLI that supports three types of operations:

• Skill management: List/view skills, install agent skills (e.g., geo-skill skills list, geo-skill install --agent codex --all)
• Website audit: Check site GEO readiness (e.g., geo-skill audit ./public, output includes passed/warning items)
• Asset generation: Generate files like robots.txt and llms.txt (e.g., geo-skill generate robots --domain https://example.com)

geo-skill natively supports mainstream AI coding agents:

• Hermes skills: Stored in skills/hermes/<skill-name>/SKILL.md, installed to ~/.hermes/skills/geo/
• Claude skills: Stored in .claude/skills/*.md, installed to ~/.claude/skills/
• Codex skills: Stored in .agents/skills/<skill-name>/SKILL.md, following OpenAI specifications This design enables seamless switching between different agents without rewriting GEO strategies.

Core values: Lower the threshold for GEO implementation, ensure coverage completeness, support automated workflows (CI/CD integration), and multi-platform adaptation (OpenAI/ByteDance ecosystem). Future roadmap: More audit checks, JSON output support for CI gates, real-time URL audit, multilingual GEO, schema generator, and benchmark test fixtures.

geo-skill promotes the transition of the GEO field from conceptual discussion to engineering practice, and it is an essential tool for product competitiveness in the AI search era. The project is open-source under the MIT license and can be quickly installed locally for experience via pip install -e ..