add-llms-txt

name	add-llms-txt
description	Use to add LLM-friendly outputs to a Hugo site — /llms.txt and /llms-full.txt indexes plus a per-page markdown twin at <url>/index.md — generated from content so they stay in sync. Triggers "add llms.txt", "make the site agent-friendly", "markdown twin".

Add llms.txt + markdown twins to a Hugo site

Never hand-edit public/ — these outputs are generated from content.

Workflow

1. Add the output formats + assignments to the site config (paste the verbatim block from this skill — copied from a working site):

   [outputs]
     home = ["HTML", "RSS", "JSON", "llms", "llmsfull"]
     # Every content page also emits a clean markdown twin at <url>/index.md for
     # LLMs/agents that fetch a single page (rendered by layouts/_default/single.markdown.md).
     page = ["HTML", "markdown"]
   
   # Dedicated media type so both LLM output formats render with a .txt extension
   # (markdown's own media type would force .md). The two output formats share it,
   # differentiated by baseName.
   [mediaTypes]
     [mediaTypes."text/llms"]
       suffixes = ["txt"]
       delimiter = "."
   
   [outputFormats]
     [outputFormats.llms]
       mediaType = "text/llms"
       baseName = "llms"
       isPlainText = true       # do not HTML-escape the markdown body
       rel = "alternate"        # head.html auto-emits a <link rel="alternate"> for it
     [outputFormats.llmsfull]
       mediaType = "text/llms"
       baseName = "llms-full"
       isPlainText = true
       rel = "alternate"
     # Per-page markdown twin. Uses Hugo's built-in text/markdown media type (.md);
     # baseName "index" yields <page-dir>/index.md. rel="alternate" makes head.html
     # advertise it on each page.
     [outputFormats.markdown]
       mediaType = "text/markdown"
       baseName = "index"
       isPlainText = true
       rel = "alternate"
   

2. Copy the templates from this skill dir into the site's layouts/:

   • index.llms.txt → layouts/index.llms.txt
   • index.llmsfull.txt → layouts/index.llmsfull.txt
   • single.markdown.md → layouts/_default/single.markdown.md

Adjust section names ("posts", "talks") to the target site's content sections. If the site has no canonical-URL field, the canonicalURL references are no-ops (.Params.canonicalURL returns empty string).

3. Customize index.llms.txt for the target site:

   • Set params.llmsIntro in hugo.toml/params.toml to a one-sentence site description for LLM crawlers — e.g. llmsIntro = "Technical writing on Kubernetes and platform engineering.". If unset, the template falls back to "Content by <author.name>.".
   • Set params.author.bio for the About line, or edit the About section directly.

4. Advertise the twin: ensure the head partial emits <link rel="alternate" type="text/markdown" href="index.md">. Congo does this automatically when rel = "alternate" is set on the output format.

Guardrails

• These outputs are generated — regenerate by rebuilding; never edit public/.
• Allow AI crawlers in robots.txt if you want them fetched (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, etc.).
• The llms-full.txt template only iterates "posts" — if the target site uses a different section name, update the range filter.

Output

/llms.txt, /llms-full.txt, and per-page index.md twins, all build-generated.