generate-site

name	generate-site
description	Orchestrate full documentation site generation for the skills-registry — structural pages from registry.yaml plus AI-enriched content and presentation-quality SVG diagrams for all plugins. Use when the user wants to rebuild the docs, update the website, regenerate pages after changing registry.yaml, add diagrams for a new plugin, or says things like "rebuild the site", "update the docs", "regenerate the website", or "I just added a plugin, update the site".
user-invocable	true
allowed-tools	["Bash","Read","Write","Edit","Agent","Skill","Glob"]

generate-site

Generate the complete documentation site for the skills-registry. Enriches each plugin with detailed content extracted from SKILL.md files, generates presentation-quality SVG diagrams, and produces structural pages from registry.yaml.

Arguments

• --plugin <name> — Only process a specific plugin. Without this, all plugins are processed.
• --no-diagrams — Skip diagram generation (enrichment only). Useful for refreshing descriptions, arguments, and argument hints without the slow diagram pipeline.

Instructions

1. Enrich plugins

Invoke /analyze-plugin for each plugin in the registry (or just the --plugin if specified).

/analyze-plugin handles every source type (github, git, git-subdir) and both plain plugins and meta-plugins (bundles) — an entry with includes is a bundle with no skills of its own, and its members are their own registry entries. So fan it over all plugins as usual, including bundles and their git-subdir members; each member gets its own enriched page, and the bundle gets an overview. (Members of one monorepo each shallow-clone it into their own .tmp/skill-repos/<name> — redundant but parallel-safe.)

If --no-diagrams is set, pass it through to /analyze-plugin so it skips Steps 5-6 (pipeline and individual skill diagrams) and only generates the enrichment YAML.

For efficiency, use sub-agents to analyze plugins in parallel (up to 6-8 at a time).

Agent({
  description: "Analyze <plugin-name>",
  prompt: "/analyze-plugin <plugin-name> [--no-diagrams]",
  run_in_background: true
})

Launch all plugin agents at once (or in batches of 6-8 for plugins with many skills). Wait for all to complete before proceeding.

Permissions

Sub-agents need these permissions to work in the background without interactive approval. Verify these are in .claude/settings.json before launching:

{
  "permissions": {
    "allow": [
      "Read(.tmp/skill-repos/**)",
      "Write(site/docs/plugins/**)",
      "Edit(site/docs/plugins/**)",
      "Bash(python3 *)",
      "Bash(d2 *)",
      "Bash(mkdir *)",
      "Bash(mv *)",
      "Bash(find *)",
      "Bash(rm *)",
      "Bash(git clone *)",
      "Bash(git pull *)",
      "Bash(open *)",
      "Bash(ls *)",
      "Skill(diagram-skills:skill-diagram:*)",
      "Skill(diagram-skills:diagram-layout:*)"
    ]
  }
}

If permissions are missing, warn the user and suggest adding them before proceeding — otherwise every sub-agent will be denied and the enrichment pass will fail silently.

2. Clean up after enrichment

After all enrichment agents complete, clean up stray files produced by the diagram-layout pipeline:

# Rename .drawio.svg → .svg (draw.io export naming convention)
find site/docs/plugins -name "*.drawio.svg" -exec sh -c 'mv "$1" "${1%.drawio.svg}.svg"' _ {} \;

# Remove temporary files
find site/docs/plugins -name "*.drawio.png" -delete
find site/docs/plugins -name "layout-plan.json" -delete

3. Generate pages

After cleanup, run the generation script again to merge enrichment data into pages:

python3 scripts/generate_site.py

This picks up _enriched.yaml files (including argument_hint and arguments) and conditionally includes SVG diagram references in pages where the SVG exists.

4. Build and verify

Stop any running dev server first (it holds a file lock):

lsof -ti:8000 | xargs kill 2>/dev/null

Build the site to verify everything works:

mkdocs build --config-file site/mkdocs.yml --site-dir /tmp/skills-site-verify-$$

Check the build output for warnings or errors. Zero warnings is the target — any "unrecognized relative link" or "target not found" warnings indicate broken references.

5. Report

Report:

• Total pages generated
• Plugins enriched
• Diagrams generated (if not --no-diagrams) — pipeline + individual count
• Build status (success/warnings/errors)
• Any skipped skills or failed diagram exports

Suggest next steps:

• Preview: mkdocs serve --config-file site/mkdocs.yml (note: new SVG files require a server restart — MkDocs doesn't hot-reload non-markdown assets)
• Deploy: push to main to trigger the GitHub Pages workflow