Refactor: Restructure skills following skill-creator best practices

- Move reference docs to references/ subdirectories (progressive disclosure)
- Slim SKILL.md files to <150 lines each (was 859 total for generator)
- Slim CLAUDE.md from 217 to 73 lines (points to skills, no duplication)
- Rename skill.md to SKILL.md for consistency
- Update the-daily-publish for Observable Framework workflow
- Add new reference files: french.md, tables.md, troubleshooting.md
- Add data dimension checklist and exploration patterns
- Add highlights selection and provincial newsworthiness guidance

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: dshkol

.claude/skills/the-daily-discover/SKILL.md

@@ -0,0 +1,101 @@
+---
+name: the-daily-discover
+description: Discover and rank StatCan CANSIM tables for Daily article generation. Use when asked to find new topics, scan for recent releases, identify coverage gaps, prioritize which tables to cover, or explore what data is available.
+---
+
+# The D-AI-LY Topic Discovery
+
+Discover and prioritize StatCan CANSIM tables for article generation.
+
+## Quick Start
+
+```bash
+# Run discovery scan
+Rscript r-tools/discover_topics.R
+
+# Check recently updated tables
+Rscript r-tools/discover_stories.R
+```
+
+## Discovery Workflow
+
+### 1. Scan Available Data
+
+```r
+library(cansim)
+library(dplyr)
+
+cubes <- list_cansim_cubes()
+
+# Filter for recent monthly releases
+candidates <- cubes %>%
+  filter(frequencyCode == "6") %>%
+  filter(cubeEndDate >= Sys.Date() - 60) %>%
+  arrange(desc(cubeEndDate))
+```
+
+### 2. Score Candidates
+
+| Dimension | Weight | Criteria |
+|-----------|--------|----------|
+| Recency | 25% | Days since release (fresher = higher) |
+| Diversity | 25% | Sector gap from existing articles |
+| Narrative | 25% | Regional variation, trend reversals |
+| Public Interest | 15% | Topic relevance to general audiences |
+| Data Quality | 10% | Complete coverage, national totals |
+
+### 3. Validate Top Candidates
+
+Before committing:
+1. Fetch sample data to verify structure
+2. Check for StatCan Daily release
+3. Verify national totals exist
+4. Confirm table is current (not deprecated)
+
+## Check Existing Coverage
+
+```bash
+ls docs/en/*/index.md | head -20
+```
+
+Avoid redundancy:
+- If CPI just published, deprioritize other price indices
+- Retail Trade + Wholesale Trade complement each other
+- Housing Starts + Building Permits complement each other
+
+## Output Format
+
+```
+TOPIC DISCOVERY RESULTS - 2025-12-24
+=====================================
+
+RANK  SCORE  TABLE        SECTOR      TITLE
+----  -----  -----------  ----------  ----------------------------------
+1     87     23-10-0079   Transport   Airline operating statistics
+2     82     18-10-0205   Prices      New Housing Price Index
+3     78     20-10-0003   Trade       Wholesale trade
+
+Top recommendation: 23-10-0079
+- Last release: 2025-12-23
+- Narrative hook: Transborder traffic down 9th consecutive month
+- Sector gap: No transport coverage in last 30 days
+```
+
+## Reference Files
+
+| Reference | When to Use |
+|-----------|-------------|
+| [sectors.md](references/sectors.md) | Sector categories, regional story patterns, scoring boosts |
+
+## Handoff to Generator
+
+After discovery, generate the article:
+
+```
+/the-daily-generator 23-10-0079 --slug airline-passengers-october-2025
+```
+
+For regional stories:
+```
+/the-daily-generator 34-10-0158 --slug ontario-housing-starts-november-2025 --geo Ontario
+```

.claude/skills/the-daily-discover/references/sectors.md

@@ -0,0 +1,103 @@
+# Sector Categories and Regional Stories
+
+Reference material for topic discovery scoring.
+
+## Sector Categories
+
+For diversity scoring, tables are categorized:
+
+| Sector | Example Tables | Notes |
+|--------|----------------|-------|
+| **Prices** | 18-10-0004 (CPI), 18-10-0001 (Gas), 18-10-0205 (NHPI) | High public interest |
+| **Labour** | 14-10-0287 (LFS), 14-10-0355 (SEPH) | Core economic indicator |
+| **Trade** | 20-10-0056 (Retail), 20-10-0003 (Wholesale), 12-10-0011 (Intl) | Supply chain coverage |
+| **Housing** | 34-10-0158 (Starts), 34-10-0292 (Permits) | Housing market health |
+| **Production** | 36-10-0434 (GDP), 16-10-0048 (Manufacturing) | Output indicators |
+| **Transport** | 23-10-0079 (Aviation), 23-10-0253 (Rail) | Mobility/logistics |
+| **Finance** | 10-10-0006 (Credit), 36-10-0580 (Investment) | Financial conditions |
+| **Demographics** | 17-10-0009 (Population), 17-10-0014 (Migration) | Social indicators |
+| **Energy** | 25-10-0015 (Electricity), 25-10-0063 (Oil & Gas) | Resource production |
+
+## Narrative Potential Indicators
+
+High-scoring narratives typically have:
+
+- **Trend reversals**: "First increase since...", "Ended X-month streak"
+- **Regional divergence**: Provinces moving in opposite directions
+- **Component splits**: House vs. land, goods vs. services, domestic vs. international
+- **Milestone crossings**: Index hits new high/low, crosses round number
+- **Seasonal anomalies**: Unexpected pattern vs. typical seasonality
+
+## Geographic Levels in CANSIM
+
+| Level | Description | Story Potential |
+|-------|-------------|-----------------|
+| **Canada** | National totals | Headline indicators |
+| **Provincial/Territorial** | 13 jurisdictions | Regional divergence, provincial spotlight |
+| **CMA** | Census Metropolitan Areas | City comparisons, metro-specific trends |
+| **Economic Region** | Sub-provincial regions | Local economic conditions |
+
+## Regional Story Types
+
+### 1. Divergence Stories
+When regions move in opposite directions:
+```r
+provincial_data %>%
+  group_by(REF_DATE) %>%
+  summarise(
+    range = max(yoy_change) - min(yoy_change),
+    leader = GEO[which.max(yoy_change)],
+    laggard = GEO[which.min(yoy_change)]
+  ) %>%
+  filter(range > 5)  # >5 percentage points spread
+```
+
+### 2. Metro Spotlight
+Deep-dive on a specific CMA:
+- Toronto housing market dynamics
+- Vancouver cost of living
+- Calgary energy sector employment
+- Montreal manufacturing
+
+### 3. Provincial Rankings
+League tables comparing provinces:
+- Unemployment rates by province
+- Housing affordability index
+- Retail sales per capita
+
+### 4. Regional Outliers
+One region bucking the national trend:
+- "Saskatchewan leads provincial gains..."
+- "Atlantic Canada bucks national decline..."
+
+## Regional Story Scoring
+
+| Factor | Score Boost | Condition |
+|--------|-------------|-----------|
+| High provincial variance | +15 | Range > 5 pp |
+| Clear leader/laggard | +10 | One province dominates |
+| CMA data available | +5 | Metro-level granularity |
+| Regional trend reversal | +20 | Province bucks national trend |
+
+## Checking Geographic Coverage
+
+```r
+check_geo_coverage <- function(table_number) {
+  df <- get_cansim(table_number)
+  geos <- unique(df$GEO)
+
+  list(
+    has_provinces = any(geos %in% c("Ontario", "Quebec", "British Columbia")),
+    has_cmas = any(grepl("CMA|Toronto|Vancouver|Montreal", geos)),
+    geo_count = length(geos),
+    geo_list = head(geos, 10)
+  )
+}
+```
+
+## Example Regional Headlines
+
+- "Toronto housing starts surge while Vancouver stalls"
+- "Prairie provinces lead employment gains in November"
+- "Quebec inflation outpaces national average for 6th month"
+- "Atlantic Canada gasoline prices hit 18-month low"