scholaraio: Skill for Claude Code

Published Paper Archive And Site Generation

Use this skill only after the user clearly says the deliverable is final, audited, or ready to publish. Drafts should stay under workspace/.

Architecture

The publishing workflow has three separate layers:

• scholaraio/: upstream code and the scholaraio publish-site generator
• published/: local audited deliverable archive, git-ignored by this repository
• an optional separate GitHub Pages repository: generated static site files

Do not put user deliverables in the repository root except under published/, and do not commit published/ to the ScholarAIO repo.

Safety Gate

Before creating or changing a published archive:

1. Confirm with the user that the deliverable is final enough to archive.
2. Recommend a backup for important outputs, especially before replacing an existing archive directory.
3. Never delete source files from workspace/; publishing is a copy/archive step.

Archive Layout

Use one folder per final deliverable:

published/<YYYY-MM-DD>-<short-title>/
├── metadata.json
├── main.pdf
├── main.tex
├── images/
└── misc/

Recommended metadata.json:

{
  "title": "Full title",
  "date": "YYYY-MM-DD",
  "keywords": ["keyword"],
  "subject": "field or topic",
  "pdf_filename": "main.pdf",
  "tex_files": ["main.tex"],
  "images_dir": "images",
  "misc_dir": "misc",
  "authors": ["Author", "Claude (AI Assistant)"],
  "note": "Final audited deliverable."
}

The site generator reads published/*/metadata.json, builds a source ZIP beside each archive, and writes a static site.

Generate The Site

Explicit output directory:

scholaraio publish-site --out-dir ~/generated-report

Configured output directory:

publish:
  site_output_dir: ~/generated-report

Then:

scholaraio publish-site

Default mode copies PDFs and source ZIPs into assets/papers/, producing a self-contained site suitable for a separate GitHub Pages repository.