[docs] Add OB1 catalog generator + committed ob1-catalog.json

Ships the site-ready catalog artifact that the natejones.com /ob1
directory will load, plus the generator script and a PR gate drift
check that keeps the artifact in sync with repo content.

Pipeline:
- scripts/build_catalog.py discovers every non-template contribution
  under the seven canonical category folders, loads metadata.json,
  rewrites intra-repo links in README.md, resolves requires_skills /
  requires_primitives into forward and reverse dependency edges, and
  emits resources/ob1-catalog.json.
- .github/workflows/ob1-gate.yml now runs the generator in --check
  mode whenever a PR touches contribution content, the generator, or
  the artifact. A drift failure tells the contributor exactly which
  command to run.

Generator-time validation (runs before artifact is emitted):
- category in metadata.json matches the folder category
- requires.open_brain is one of required | optional
- learning_order is only set on extensions
- every requires_skills / requires_primitives slug resolves to a
  real contribution
- no two entries share the same site_path
- every relative intra-repo link in a contribution README resolves
  to an existing file in the repo

Link rewriting:
- relative links to contribution folders or their READMEs become
  /ob1/<category>/<slug>
- links to category index folders or READMEs become /ob1/<category>
- deeper relative files inside contribution folders go to the raw
  GitHub blob URL
- image / asset references go to raw.githubusercontent.com
- other intra-repo links (docs/, root markdown) go to blob/tree URLs
- absolute URLs, anchors, and mailto: / tel: schemes pass through

Site loader precedence (documented here for PR3 in the site repo):
- OB1_CATALOG_LOCAL_PATH env var takes precedence when set
- otherwise fetch from raw.githubusercontent.com on main
- missing or invalid artifact must fail the site build

Pre-existing fix included:
- recipes/email-history-import/README.md linked to a non-existent
  primitives/content-fingerprint-dedup path. The dedup logic lives in
  recipes/content-fingerprint-dedup. The generator's strict link
  check flagged it, so fixed the link instead of loosening the rule.

Output: 59 catalog entries across all seven categories
(recipes 29, skills 13, extensions 6, primitives 5, integrations 3,
dashboards 2, schemas 1) — 4 required + 9 optional in skills, all
other categories required for v1.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: justfinethanku

.github/workflows/ob1-gate.yml

@@ -31,6 +31,29 @@ jobs:
       - name: Install metadata schema validator
         run: python3 -m pip install check-jsonschema
 
+      - name: Check catalog is not stale
+        id: catalog_check
+        # Only run if any contribution content or metadata changed — a doc-only
+        # or workflow-only PR cannot invalidate the catalog.
+        run: |
+          set +e
+          changed=$(git diff --name-only origin/main...HEAD)
+          if echo "$changed" | grep -qE '^(recipes|schemas|dashboards|integrations|skills|primitives|extensions)/[^_].*/(README\.md|metadata\.json)$' \
+            || echo "$changed" | grep -qE '^scripts/build_catalog\.py$' \
+            || echo "$changed" | grep -qE '^resources/ob1-catalog\.json$'; then
+            python3 scripts/build_catalog.py --check
+            rc=$?
+            if [ "$rc" -ne 0 ]; then
+              echo "catalog_stale=true" >> $GITHUB_OUTPUT
+              echo "::error::resources/ob1-catalog.json is stale. Run \`python3 scripts/build_catalog.py\` and commit the result."
+              exit "$rc"
+            fi
+            echo "catalog_stale=false" >> $GITHUB_OUTPUT
+          else
+            echo "catalog check skipped (no contribution or generator changes)"
+            echo "catalog_stale=false" >> $GITHUB_OUTPUT
+          fi
+
       - name: Get changed files
         id: changed
         run: |

recipes/email-history-import/README.md

@@ -100,7 +100,7 @@ deno run --allow-net --allow-read --allow-write --allow-env pull-gmail.ts --list
 4. **Deduplicate** via sync-log (tracks Gmail message IDs already imported)
 5. **Embed** content via OpenRouter (`text-embedding-3-small`)
 6. **Classify** via LLM (topics, type, people, action items)
-7. **Upsert** into Supabase with SHA-256 [content fingerprint](../../primitives/content-fingerprint-dedup/README.md) — re-running produces zero duplicates
+7. **Upsert** into Supabase with SHA-256 [content fingerprint](../content-fingerprint-dedup/README.md) — re-running produces zero duplicates
 
 ### What gets filtered out
 
@@ -115,7 +115,7 @@ Each imported email becomes one row in the `thoughts` table:
 - `content`: Email body with context prefix (`[Email from X | Subject: Y | Date: Z]`)
 - `embedding`: 1536-dim vector for semantic search (truncated to 8K chars)
 - `metadata`: LLM-extracted topics, type, people, action items, plus `source: "gmail"`, `gmail_id`, `gmail_labels`, `gmail_thread_id`
-- `content_fingerprint`: Normalized SHA-256 hash for dedup (see [content fingerprint primitive](../../primitives/content-fingerprint-dedup/README.md))
+- `content_fingerprint`: Normalized SHA-256 hash for dedup (see [content fingerprint primitive](../content-fingerprint-dedup/README.md))
 
 ## Troubleshooting