Sync Curated Content

Pipeline skill: ingest external learning material (newsletters, articles forwarded to a dedicated alias) into the knowledge base. Runs standalone or as step 3 of /data-ingestion.

Setup convention

Specialized industry newsletters are subscribed via the alias henrygoldsmith07+newsletter@gmail.com, so they are filterable with the Gmail query to:henrygoldsmith07+newsletter@gmail.com.

Process

1. Find the last sync point. Read raw/curated/.last-sync; default to 14 days ago if missing.
2. Fetch. Use Gmail search_threads with the alias query plus after:<last-sync date>. If the Gmail connector isn't available, say so and stop — don't guess at content.
3. Extract claims, not prose. For each newsletter/article, pull:
   • Concrete claims and techniques (with the source's reasoning)
   • Tools, releases, or practices worth evaluating
   • Anything that contradicts an existing wiki entry — flag these explicitly
4. Write the digest. One file per run: raw/curated/YYYY-MM-DD_newsletter-digest.md, one section per source with sender and subject as provenance.
5. Update the wiki. Add a row to the Curated Content table in wiki/index.md. Where a claim updates an existing wiki/<topic>.md page, edit that page: add the new claim with a dated citation, and mark contradicted statements as superseded rather than deleting them.
6. Update the sync point in raw/curated/.last-sync.
7. Report. Sources read, claims extracted, wiki pages updated, contradictions flagged. If run from /data-ingestion, keep this to 3 lines.