Paper Workbench

Unified entrypoint for paper-like inputs.

Use this skill when the user provides any of these and wants structured paper intake or downstream analysis:

• arXiv IDs and arXiv URLs
• AlphaXiv URLs
• DOI strings or doi.org/... URLs
• local academic PDFs
• remote PDF URLs
• paper landing pages that expose a PDF
• existing normalized paper JSON

This skill is a thin router. Always normalize first into paper-record, then choose the requested mode.

Routing Rule

1. Resolve source from $ARGUMENTS, latest user message, or pasted JSON
2. Normalize via the bundled scripts/normalize_paper.py
3. Route by mode:
   • json → return canonical JSON record
   • interpret → explain the paper using normalized facts
   • xray → deconstruct the paper using normalized facts

If mode is omitted, default to json.

Normalize First

Always run:

python "$SKILL_DIR/scripts/normalize_paper.py" \
  --source "<paper-source>" \
  --lang "<lang>" \
  --fulltext auto

If the user wants saving, pass --save only when saving the JSON artifact. Interpret and xray save behavior stays owned by their downstream skills.

Mode Handoff

• For json, return the paper-record payload directly
• For interpret, treat the normalized JSON as the fact source and follow references/modes/interpret.md
• For xray, treat the normalized JSON as the fact source and follow references/modes/xray.md

References

• references/routing.md — source classification and mode routing
• references/schema.md — canonical paper-record contract
• references/migration.md — legacy compatibility rules
• references/modes/json.md — JSON mode behavior
• references/modes/interpret.md — interpretation handoff
• references/modes/xray.md — x-ray handoff