feat(site): publish the catalogue on GitHub Pages

.github/workflows/pages.yml

@@ -0,0 +1,34 @@
+name: pages
+# Publishes site/ (the catalogue at https://stylusnexus.github.io/agent-plugins/)
+# to GitHub Pages whenever it changes on main. ci.yml's check-readme.py keeps the
+# catalogue's packs, skills, and counts in step with plugins/ on disk.
+on:
+  push:
+    branches: [main]
+    paths: ["site/**", ".github/workflows/pages.yml"]
+  workflow_dispatch:
+
+permissions:
+  contents: read
+  pages: write
+  id-token: write
+
+concurrency:
+  group: pages
+  cancel-in-progress: false
+
+jobs:
+  deploy:
+    name: deploy
+    runs-on: ubuntu-latest
+    environment:
+      name: github-pages
+      url: ${{ steps.deployment.outputs.page_url }}
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/configure-pages@v6
+      - uses: actions/upload-pages-artifact@v5
+        with:
+          path: site
+      - id: deployment
+        uses: actions/deploy-pages@v5

README.md

@@ -17,7 +17,7 @@ Windsurf, Zed, opencode, Cline, Continue, Hermes and others) through the Skills
 Every skill is repo-agnostic: it detects your repository's conventions rather than assuming its own,
 and defers to a repo-local version of itself when your project defines one.
 
-**[Browse the catalogue →](https://claude.ai/code/artifact/667c2158-4833-4535-a5c5-a5597dea8fea)** — every pack, what it's for, and how to install it on your agent.
+**[Browse the catalogue →](https://stylusnexus.github.io/agent-plugins/)** — every pack, what it's for, and how to install it on your agent.
 
 > ⚠️ **Trust before you install.** Plugins run code on your machine. Review a plugin's source before
 > installing it. Everything here is open-source — links are in the table below.
@@ -484,6 +484,7 @@ agent-plugins/
 │   ├── gen-catalog.py           # writes the skill catalogue and skills.sh.json
 │   ├── check-links.py           # relative links and anchors resolve
 │   └── check-private-terms.sh   # nothing private ships in a skill
+├── site/                        # the catalogue page, published to GitHub Pages
 ├── skills.sh.json               # skills.sh groupings, one per pack
 ├── LICENSE
 └── README.md

package.json

@@ -13,7 +13,7 @@
     "github-copilot",
     "developer-tools"
   ],
-  "homepage": "https://github.com/stylusnexus/agent-plugins#readme",
+  "homepage": "https://stylusnexus.github.io/agent-plugins/",
   "bugs": {
     "url": "https://github.com/stylusnexus/agent-plugins/issues"
   },

scripts/check-private-terms.sh

@@ -5,6 +5,9 @@
 # has no business in this PUBLIC MIT marketplace's shipped text.
 #
 # Scope is plugins/*/skills/**, plugins/*/agents/**, and plugins/*/README.md.
+# The root README.md and site/ (the GitHub Pages catalogue) are public too, but
+# they name this org's repos and publisher on purpose, so they're checked
+# against the private overlay only (failures and warnings), like docs/ below.
 # plugin.json and LICENSE are never scanned -- the publisher name there is
 # legitimate and the whole point of the LICENSE file.
 #
@@ -153,6 +156,36 @@ elif [ -d docs ]; then
   echo "no private-terms overlay available -- skipping docs/ check" >&2
 fi
 
+# Public pages outside plugins/: the root README and the Pages site. Private
+# overlay terms fail; warn: vocabulary warns. Public-list terms don't apply here
+# (these pages link this org's repos by design).
+PAGES=""
+[ -f README.md ] && PAGES="README.md"
+[ -d site ] && PAGES="$PAGES site"
+if [ -n "$PAGES" ] && [ -s "$OVERLAY_FILE" ]; then
+  while IFS= read -r pattern; do
+    case "$pattern" in
+      ''|'#'*) continue ;;
+      warn:*)
+        # shellcheck disable=SC2086
+        hits=$(grep -rniIE "${pattern#warn:}" $PAGES 2>/dev/null || true)
+        if [ -n "$hits" ]; then
+          echo "WARNING: private-product vocabulary (from the private list) found; check these lines:" >&2
+          report_warning "$hits"
+          warned=1
+        fi
+        continue ;;
+    esac
+    # shellcheck disable=SC2086
+    hits=$(grep -rniIE "$pattern" $PAGES 2>/dev/null || true)
+    if [ -n "$hits" ]; then
+      echo "PRIVATE TERM (from the private list) found in README.md or site/ at:" >&2
+      echo "$hits" | cut -d: -f1,2 | sed 's/^/  /' >&2
+      fail=1
+    fi
+  done < "$OVERLAY_FILE"
+fi
+
 [ "$warned" = 1 ] && echo "warnings above are not failures: confirm each line is generic, or rewrite it" >&2
-[ "$fail" = 0 ] && echo "no private terms found in plugins/*/skills, plugins/*/agents, plugins/*/README.md, or docs/"
+[ "$fail" = 0 ] && echo "no private terms found in plugins/*/skills, plugins/*/agents, plugins/*/README.md, docs/, README.md, or site/"
 exit "$fail"

scripts/check-readme.py

@@ -13,6 +13,9 @@
   3. Each plugins/<pack>/README.md must mention every skill under that pack's
      skills/ directory (found as a table cell that is exactly one backticked
      name), and mention no skill that doesn't exist.
+  4. The GitHub Pages catalogue (site/index.html) must have a card for every
+     pack with the right skill count and one chip per skill, and its headline
+     plugin and skill counts must match the README's.
 
 The table format is read from the README rather than hardcoded (look for the
 header row whose first cell is "Pack" / "Plugin"), so formatting can move
@@ -219,15 +222,58 @@ def check_plugin_readmes():
             errors.append(f"{readme}: table(s) mention skill(s) not in this pack: {sorted(extra)}")
 
 
+SITE_INDEX = os.path.join("site", "index.html")
+
+
+def check_site():
+    if not os.path.isfile(SITE_INDEX):
+        return
+    with open(SITE_INDEX, encoding="utf-8") as fh:
+        html = fh.read()
+    cards = {}
+    for art in re.findall(r'<article class="plugin">(.*?)</article>', html, re.S):
+        m = re.search(r'class="plugin-name">(?:<a [^>]*>)?([a-z0-9-]+)', art)
+        if m:
+            cards[m.group(1)] = art
+    total = 0
+    for pack in sorted(os.listdir(PLUGINS_DIR)):
+        real = actual_skills(pack)
+        if not real:
+            continue
+        total += len(real)
+        art = cards.get(pack)
+        if art is None:
+            errors.append(f"{SITE_INDEX}: no card for pack '{pack}'")
+            continue
+        n = re.search(r'<span class="plugin-n">(\d+) skills', art)
+        if not n or int(n.group(1)) != len(real):
+            errors.append(f"{SITE_INDEX}: card '{pack}' should say {len(real)} skills")
+        chips = set(re.findall(r'<span class="chip">([a-z0-9-]+)</span>', art))
+        if chips != set(real):
+            errors.append(
+                f"{SITE_INDEX}: card '{pack}' chips differ from plugins/{pack}/skills/: "
+                f"missing {sorted(set(real) - chips)}, extra {sorted(chips - set(real))}"
+            )
+    with open(ROOT_README, encoding="utf-8") as fh:
+        headline = re.search(r'\*\*(\d+)\s+plugins?,\s*(\d+)\s+skills?\*\*', fh.read())
+    site_plugins = re.search(r'<b>(\d+)</b><span>plugins</span>', html)
+    site_skills = re.search(r'<b>(\d+)</b><span>skills</span>', html)
+    if not site_skills or int(site_skills.group(1)) != total:
+        errors.append(f"{SITE_INDEX}: headline skill count should be {total}")
+    if headline and (not site_plugins or site_plugins.group(1) != headline.group(1)):
+        errors.append(f"{SITE_INDEX}: headline plugin count should be {headline.group(1)}")
+
+
 def main():
     check_root_readme()
     check_plugin_readmes()
+    check_site()
     if errors:
         print(f"{len(errors)} README/skill-set mismatch(es) found:\n", file=sys.stderr)
         for e in errors:
             print(f"  {e}", file=sys.stderr)
         return 1
-    print("README pack tables, plugin READMEs, and headline counts all match plugins/ on disk")
+    print("README pack tables, plugin READMEs, headline counts, and the Pages catalogue all match plugins/ on disk")
     return 0