new-papers

When to Use This Skill

Use this skill to discover new papers published since your last scan in categories relevant to cortical column research.

Trigger phrases:

• "What's new since last time?"
• "Scan for new papers"
• "Check for new publications"
• "Any new papers in my field?"
• "Quoi de neuf dans la littérature ?"

Workflow

Step 1: Check Last Scan Timestamp

1. Look for ~/Research/cortical-columns/04 - Literature Map/last-scan.md
2. If it exists, read the timestamp to determine days since last scan
3. If it doesn't exist (first run): inform the user "Première analyse — je vais scanner les 7 derniers jours." and use 7 days as default

Step 2: Ensure Obsidian Vault Exists

Check if ~/Research/cortical-columns/ exists. If not:

1. Create the vault structure at ~/Research/cortical-columns/ with all required folders
2. Create all folders: 00 - Inbox/, 01 - Foundational Papers/, 02 - Recent Discoveries/, 03 - Concepts/, 04 - Literature Map/, 05 - Thesis Notes/
3. Confirm vault creation to the user in French: "Votre vault Obsidian a été créé avec succès."

Step 3: Search arXiv for Recent Papers

For each relevant arXiv category, use Claude's web search to find recent papers:

1. Search arxiv.org for recent papers in each category (cs.AI, cs.NE, q-bio.NC, cs.LG, cs.CL, stat.ML)
2. Use web search with category-specific queries like "arxiv cs.AI new papers this week" or fetch from https://arxiv.org/list/{category}/recent
3. Combine all results, remove duplicates by arXiv ID

Step 4: Quick Scan & Relevance Scoring

For each unique paper:

1. Evaluate relevance directly using Claude's AI capabilities
2. Score each paper 1-10 based on relevance to cortical column research
3. Filter: keep only papers with score >= 7

Step 5: Check for Duplicates in Vault

Before creating any note:

1. Scan 00 - Inbox/, 01 - Foundational Papers/, and 02 - Recent Discoveries/ for existing notes
2. Check by arXiv ID in the YAML frontmatter
3. If duplicate found: skip creation, note that paper already exists

Step 6: Create Obsidian Notes

For each unique, relevant paper (score >= 7):

1. Create a markdown file in ~/Research/cortical-columns/02 - Recent Discoveries/
2. Filename format: {arxiv_id}-{slugified-title}.md
3. Use the quick-scan template format:
   • YAML frontmatter with: title, authors, year, arxiv_id, source, relevance_score, status, tags, date_added
   • Abstract section
   • Quick Assessment section (key contribution, relevance to thesis, should deep-dive)
   • Related Papers section with Obsidian backlinks
4. Add a "Source: periodic scan" tag

Step 7: Update Last Scan Timestamp

1. Create or update ~/Research/cortical-columns/04 - Literature Map/last-scan.md
2. Content:

# Last Scan

**Timestamp**: {current ISO timestamp}
**Categories scanned**: cs.AI, cs.NE, q-bio.NC, cs.LG, cs.CL, stat.ML
**Results**: {count} new papers found

Step 8: Output Summary to User

Present findings with a French summary (résumé en français):

• "J'ai trouvé X nouveaux articles pertinents depuis le {last_scan_date}."
• List each paper with: title, category, relevance score, and brief reason
• If no new papers: "Aucun nouvel article pertinent depuis votre dernière analyse."

Error Handling

• arXiv search unavailable: "L'API arXiv n'est pas disponible actuellement."
• No new papers: "Aucun nouvel article pertinent trouvé dans les catégories scannées."
• First run: Use 7-day default, inform user

Relevant arXiv Categories

• cs.AI: Artificial Intelligence
• cs.NE: Neural and Evolutionary Computing
• q-bio.NC: Neurons and Cognition (computational neuroscience)
• cs.LG: Machine Learning
• cs.CL: Computation and Language (NLP)
• stat.ML: Machine Learning (statistics)

Output Format

The skill outputs:

1. Markdown notes in 02 - Recent Discoveries/ folder
2. Updated last-scan timestamp in 04 - Literature Map/last-scan.md
3. A French summary message to the user