Librarian ingest skill

Goal

• •Add a new source document to the vault in a reproducible, file-based way.
• •For owned material: keep raw + clean Markdown transcript in-repo.
• •For reference-only material: store metadata + notes only; do NOT commit full text.

Inputs

• •A local file path or a URL to a PDF/HTML/TXT.
• •License status: owned or reference-only.
• •Optional identifiers: DOI, arXiv, URN, canonical URL.

Outputs

• •vault/raw/<doc_id>/original.<ext>
• •vault/owned/<doc_id>/doc.md (+ transcript.md) OR vault/references/<doc_id>/doc.md
• •Updated catalog/*.

Procedure

1. •

   Generate a document id:

   • •python tools/townctl.py id doc
2. •

   Ingest:

   • •python tools/townctl.py ingest --source <file> --license-status owned --title "<title>" --url "<url>" --doi "<doi>" --urn "<urn>"
   • •For reference-only, use --license-status reference-only.
3. •

   Regenerate catalog and site exports:

   • •python tools/townctl.py index
   • •python tools/townctl.py export-site
4. •

   Validate:

   • •python tools/townctl.py lint

Guardrails

• •Never commit full text for reference-only documents.
• •Put anything non-owned/sensitive into vault/private/ (gitignored) and commit only metadata + your own notes.