Monitor Blogs and Feeds with Hermes Blogwatcher

Blogwatcher provides a durable subscription and scan workflow for blogs and feeds. It can discover RSS or Atom sources, fall back to scraping where supported, store scan state in SQLite, list unread articles, and mark items as read. Container use requires persistent storage or every run starts from an empty history.

Official skill reference: Blogwatcher

Separate collection from publication

Adding a blog and scanning it only creates a monitoring queue. New entries should still be deduplicated, opened, checked for source identity and date, and reviewed before becoming a summary, alert, or public post. A feed title is metadata, not proof of the article's claims.

For Docker, mount a named volume or host directory and point the database environment setting at that persistent location. When migrating from an earlier Blogwatcher database, make a backup and verify the new CLI recognizes existing subscriptions and article state.

Monitoring checklist

Confirm the canonical site and discovered feed URL.
Keep the SQLite database on persistent storage.
Run one baseline scan before treating later items as new.
Review unread items before marking them read.
Deduplicate cross-posts and redirected URLs.
Record source, author, publication date, and original link in downstream work.

Common pitfalls

Running a disposable container without a volume. Scan history and subscriptions vanish.
Publishing feed text automatically. Feeds may be truncated, stale, malformed, or copyrighted.
Marking everything read before a successful handoff. Failed downstream processing can then lose visibility.
Treating scraping fallback as guaranteed. Site changes, blocks, and JavaScript rendering can interrupt collection.
Scanning too aggressively. Respect site resources and use a sensible cadence.

Verification

List tracked blogs, run a scan, and inspect the unread article list. Confirm a known item contains the expected title, canonical URL, source, and date. Restart the container or process once to prove state persists. After any downstream publication, compare the public attribution and link with the original source before marking the item complete.