AI Weekly Newsletter Skill

Generate curated weekly newsletter from stories marked [x] in daily AI digests.

Prerequisite: User must check - [x] on desired stories in daily digest files (Notion renders these as clickable checkboxes).

Arguments

Parse from $ARGUMENTS:

• --week YYYY-WNN — Target week (e.g., 2026-W06). Default: current week
• --dry-run — Show extraction results without saving
• --force — Allow re-publishing a week already in .published-issues

State Files

Published Issues

Track published newsletters in:

./findings/ai-newsletter/.published-issues

Format: YYYY-WNN|YYYY-MM-DD (one line per issue)

Example:

2026-W05|2026-02-02
2026-W06|2026-02-09

Workflow

Phase 1: Setup

1. Parse arguments for --week, --dry-run, --force
2. Determine target week:
   • If --week provided: use that week
   • Default: current ISO week
3. Calculate date range (Monday to Sunday of target week)
4. Read .published-issues file:
   • If target week already published AND no --force flag → error with message: "Week {YYYY-WNN} already published on {date}. Use --force to regenerate."
   • If missing: create empty file
5. Read output-template.md for newsletter format

Phase 2: Extract Checked Stories

1. Glob digest files in date range:

./findings/ai-daily-digest/ai-digest-{YYYY-MM-DD}.md

Match files where date falls within Mon-Sun of target week.

2. For each digest file:

   • Read file content
   • Extract all lines matching - [x] ** (checked story items)
   • For each matched line, capture:
     • title — text between ** markers
     • summary — text after ** and before [[Source] or [Source
     • url — source URL
     • section — nearest ## or ### header above the line
     • date — from filename

3. If zero checked stories found:

   • Error: "No stories marked [x] found in digests for week {YYYY-WNN} ({date_range}). Check stories in Notion first."
   • Abort

4. Store extracted stories with metadata.

Phase 3: Categorize into Themes

Group stories into narrative themes based on content similarity:

Rules:

• 1-2 stories total: Skip theming, use simple list format
• 3-7 stories: Group into 2 themes
• 8-15 stories: Group into 2-4 themes + Quick Hits overflow
• 16+ stories: 3-4 themes + generous Quick Hits

Theme generation:

• Analyze story titles, summaries, and sections
• Create narrative theme titles (NOT mechanical section names)
• Pick appropriate emoji for each theme
• Select lead story for each theme (most impactful)
• Overflow stories (don't fit themes cleanly) → Quick Hits

Good theme names:

• "The Open Source Surge" (not "Open Source AI")
• "Agents Are Everywhere" (not "AI Tools")
• "The Infrastructure Race" (not "Hardware & Infra")
• "When AI Meets Science" (not "Research Papers")

Bad theme names:

• Direct copies of digest section names
• Single-word themes
• Generic "AI News" / "Updates"

Phase 4: Generate Newsletter

1. Load output-template.md
2. Generate content:

Editorial opening (This Week in AI):

• 2-3 sentences capturing the week's narrative
• Reference the biggest story
• Set the tone (exciting, concerning, transformative, etc.)

Themed sections:

• 1-2 sentence context for the theme
• Each story: enhanced summary (expand beyond original 1-liner)
• Append | *{Mon DD}* date to each story
• "Why this matters" analysis after each theme's stories

Quick Hits:

• 1-line per story (title + brief summary + source)
• For overflow that didn't fit themes

Looking Ahead:

• 2-3 bullet points on trends to watch based on this week's stories
• Forward-looking, not summary

3. If --dry-run: output to console, skip Phase 5

Phase 5: Save Newsletter

Step 1: Create directory if needed

mkdir -p ./findings/ai-newsletter

Step 2: Write newsletter to Notion workspace

Use Notion MCP to create page in workspace with title AI Newsletter {YYYY-WNN} and content formatted per template.

Step 3: Write archive copy

./findings/ai-newsletter/ai-newsletter-{YYYY-WNN}.md

Step 4: Update .published-issues

Append line: {YYYY-WNN}|{today's date YYYY-MM-DD}

Verification: Confirm Notion page created + 1 archive file + 1 state file updated.

Output Requirements

• Use emojis in section headers
• Bullet points over paragraphs
• Every story must have source URL
• Include story count and digest count in header
• Wikilinks to vault notes where relevant
• Tags: #ai-newsletter #weekly #YYYY-WNN

Error Handling

• Zero marked stories → clear error message, abort
• Missing digest files for date range → warn, continue with available files
• Already published week → blocked unless --force
• Single digest file in range → proceed normally (partial week OK)

Example Invocations

# Current week newsletter
/ai-newsletter

# Specific week
/ai-newsletter --week 2026-W06

# Preview without saving
/ai-newsletter --dry-run

# Regenerate already-published week
/ai-newsletter --week 2026-W05 --force