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.

4.0 KB
Raw Blame

Source Map

Raw Sources

  • .raw/sources/claude-blog-brain-data/google-updates.json: immutable historical snapshot of the repo's Google update ledger captured 2026-07-09. It is provenance, not the current ledger.
  • The live claude-blog skill surface (skills/, agents/, scripts/, and plugin metadata at the repository root) is the current reference; the Brain keeps no raw copy of it.
  • Google Search Central docs; web.dev CWV; FLOW framework bibliography; GEO/AEO studies; content-marketing and copywriting research.

Enrichment Sources

  • Official Google Search Central, web.dev, and Schema.org documentation
  • Primary-source-verified Google algorithm-update ledger (data/google-updates.json)
  • Quality Rater Guidelines (E-E-A-T, YMYL, scaled-content abuse) and the Search Essentials / spam policies
  • FLOW framework (github.com/AgriciDaniel/flow) and its cited studies
  • GEO/AEO and AI-search studies (AI Overviews coverage and CTR, AI Mode behavior, passage-level citability)

Import Strategy

  • Copy raw source files into .raw/sources/.
  • Record path, hash, retrieval date, owner, and source type.
  • Use sha256 only and keep all raw paths vault-relative.
  • Record external research sources in references/source-ledger.json.
  • Give each record an explicit lifecycle: active, retired or unverified. Only active, reviewed, unexpired evidence can support current recommendations. Retired records retain their historical dates and name active replacement IDs and a retirement reason; do not fetch or refresh them as current evidence.
  • For active records, review the claim scope against content, record the actual retrieval and review dates separately from publication/update dates, and keep limitations visible. Record the excerpt, rationale, normalized SHA-256 and safe repository-relative full-capture path in verification. Also record captured_excerpt_path and captured_excerpt_sha256 for the packaged reviewed excerpt artifact under references/evidence/. Availability and token overlap are diagnostics, not claim verification.
  • verify_source_ledger.py --offline-check requires actual regular, nonsymlinked excerpt artifacts inside that fixed evidence root. It checks file hashes, normalized excerpt agreement, source identity, URL, review and retrieval dates, disposition and the recorded full-document hash binding. The aggregate redistributed excerpts for each document are capped at 25 words. The offline gate does not recompute the full-document digest, prove full-page availability or determine semantic entailment. Preserve the separate full-capture pack and independently check its hashes and excerpts. Secure descriptor-relative reads are verified on Linux and fail closed on unsupported platforms. Coverage is the unique paths referenced by active sources, with every artifact record and excerpt bound to the active ledger. Unreferenced and historical files are outside this gate. A new retrieval attempt must not overwrite a successful review until its claims are reviewed.
  • When a ledger entry relies on a captured raw file, record raw_snapshot_path and raw_snapshot_sha256.
  • Record implemented schemas and adapters in references/adapter-manifest.json.
  • Treat repository-root data/google-updates.json as canonical. Generate brain/data/google-updates.json with scripts/sync_google_updates.py; never edit the Brain projection independently.
  • Create a source note under wiki/sources/.
  • Link affected entities, workflows, and deliverables.

Source Note Convention

  • Source-note IDs must match references/source-ledger.json source IDs.
  • Source notes use wiki/sources/<source-id>.md with a backlink to the ledger ID.
  • Each source note must link at least one affected workflow, claim, canon note, or deliverable.
  • Do not add source-only prose without a matching ledger entry.
  • Raw-source paths must be vault-relative normalized paths under .raw/sources/, never absolute paths and never symlink exits.