Feat/llms txt generator

[kuibinlin]

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Improvement (refactoring and improving code)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update

Description

Movtivation:

SEO helps static sites rank on Google, but it does nothing for AI-powered search tools like ChatGPT, Perplexity, and Claude, which are increasingly how people discover and summarize personal blogs.

These tools have no structured way to understand a Jekyll (Chirpy theme) site. They parse raw HTML through navigation, sidebars, and footers to piece together what the site is about. Blog owners who put effort into writing content have no way to optimise their visibility in LLM-powered search results.

Just as sitemap.xml solved discoverability for traditional search engines, llms.txt (https://llmstxt.org/) is an emerging standard that solves it for LLMs — giving Chirpy theme users a way to go beyond traditional SEO and make their sites visible to the next generation of search.

Solution:

A generator plugin _plugins/llms-txt-generator.rb that creates /llms.txt (and optionally /llms-full.txt) during the Jekyll build, following the llmstxt.org spec.

llms.txt — a clean markdown index with the site title, description, and links to all pages and posts
llms-full.txt — the full body content of every page and post in one file, so an LLM can understand the entire site in a single request
This would help Chirpy theme users' blogs surface better when people ask AI tools questions related to their content.

Implementation details:

• Opt-in via llms_txt.enabled: true in _config.yml
• Zero impact when disabled
• Auto-reads existing _data/contact.yml for a contacts section
• No new dependencies (Nokogiri already ships with Jekyll)

Additional context

Fixes feat: generate /llms.txt for LLM-friendly site content #2690

Fixed feat: add llms.txt generator plugin for LLM-friendly content #2692

Fixed feat: generate /llms.txt and /llms-full.txt for LLM-friendly site content #2693

Here's why this is the better fit:

• Spec compliance — llms-full.txt is supposed to contain the full page/post content so LLMs can consume it without crawling the site. This version inlines the complete rendered text (HTML-stripped). The other PR feat: generate /llms.txt and /llms-full.txt for LLM-friendly site content #2691 feat: add llms.txt generator plugin for LLM-friendly content #2692 feat: generate /llms.txt and /llms-full.txt for LLM-friendly site content #2693 only includes excerpts/descriptions, which makes the full variant essentially redundant with llms.txt.

• Opt-in by default — This version requires enabled: true to activate. That's safer behavior for a plugin that generates new public-facing files. The other PR feat: generate /llms.txt and /llms-full.txt for LLM-friendly site content #2691 feat: add llms.txt generator plugin for LLM-friendly content #2692 feat: generate /llms.txt and /llms-full.txt for LLM-friendly site content #2693 generates files automatically on install, which could surprise users.

• Contacts section — This version auto-resolves contacts from _data/contact.yml with smart URL inference for GitHub, Twitter, email, and RSS.

• Optional links section — Per the spec, there's an "Optional" section for supplementary links that LLMs can skip when context is limited. This version supports it; the other doesn't.
  Configuration — Supports custom title, description, and full-text toggle, all clearly documented with inline comments.

• Nokogiri for HTML stripping — Uses Nokogiri instead of regex, which handles edge cases better. Nokogiri is already a Jekyll dependency so there's no added overhead.