Skip to content

Free, open source blog writer for Claude Code. Researches current sources, writes answer-first posts for Google and AI answers, has an independent reviewer score each one out of 100 and ships only after five delivery gates. 32 skills, 30 /blog commands, 5 agents. Active development at AI-Marketing-Hub/claude-blog; public releases ship here.

2.2 KB
Raw Blame

Claude Blog Brain Operator Kit

Five-Minute Path

python -m pip install -e .
claude-blog-brain demo
claude-blog-brain lint --vault examples/sample-vault
claude-blog-brain report --vault examples/sample-vault --html-only

Open examples/sample-vault/ in Obsidian and read:

  1. CODEX.md
  2. wiki/hot.md
  3. wiki/index.md
  4. wiki/meta/dashboard.md

Client Vault

claude-blog-brain new acme --client-name "Acme Co" --owner "Daniel Agrici" --out-dir ~/claude-blog-brain-vaults
claude-blog-brain ingest --vault ~/claude-blog-brain-vaults/acme --file tests/fixtures/sample-source.md
claude-blog-brain synthesize --vault ~/claude-blog-brain-vaults/acme
claude-blog-brain report --vault ~/claude-blog-brain-vaults/acme --html-only

Research Rule

Refresh official or primary sources when their review windows expire or a material claim changes. The local 2026-10-07 readiness audit passes at 98/100; that result does not survive source expiry or later changes automatically.

Research evidence must be written into references/source-ledger.json with source URL, source type, published or last-updated date, retrieved date, date precision, refresh due date, confidence, evidence tier, limitations, and claim coverage. Markdown research notes alone do not satisfy market-ready release. Active records require a dated claim review, excerpt, rationale and normalized content hash. Retired and unverified records cannot support current advice. The offline ledger check requires contained, nonsymlinked reviewed excerpt artifacts under references/evidence/. It checks their bytes, hashes, excerpt agreement and source provenance. It does not recompute the retained full-document hash or establish full-page availability or semantic entailment. Retain and independently hash-check the separate full-capture evidence pack. Secure descriptor-relative excerpt reads are verified on Linux; unsupported platforms fail closed. See references/source-map.md for lifecycle and capture conventions.

Adapter Rule

Domain-adapted status requires references/adapter-manifest.json to name real schemas, importer paths, synthesis modules, report renderers, fixtures, and tests. Generic scaffold scripts are intentionally capped below market-ready.