version: "1.0.1" created: "2025-10-21 14:35" updated: "2025-10-21 14:35" system_type: "claude-code-skill" status: "active"

Daily Workflow Automation Skill

Overview

Purpose: Automatically processes daily notes when users mention diary extraction, end-of-day workflows, or ask "what did I work on today". Extracts entries from ### 🧠 Notes sections, classifies them using AI, routes them to appropriate context/diary/insights/ideas files, and updates memory systems (Serena + Graphiti).

Type: Claude Code Skill (Auto-triggered)

Status: Active (Phase 1 Implementation Complete)

Replaces: /extract-daily-content command (manual slash command)

Location

Primary Files

• .claude/skills/daily-workflow/SKILL.md - Main skill workflow (loads when triggered)
• .claude/skills/daily-workflow/reference/templates.md - Daily note template documentation
• .claude/skills/daily-workflow/reference/formatting-rules.md - Entry transformation rules
• .claude/skills/daily-workflow/scripts/extract_content.py - Deterministic extraction script

Configuration

• .claude/skills/README.md - Skill registry and testing guide
• 98-templates/daily-note.md - Daily note template with ### 🧠 Notes section
• .claude/commands/.extract-daily-content-tracker.json - Extraction tracker (auto-created)

Related Files

• .claude/commands/extract-daily-content.md - Original command (now deprecated in favor of skill)
• 00 - Daily/ - Daily notes directory (extraction source)
• 04-resources/diary.md - Primary diary destination
• 04-resources/context.md - Primary context destination
• 01-areas/p-dev/insights.md - Primary insights destination
• 04-resources/ideas.md - Primary ideas destination

Architecture

Components

1. SKILL.md (.claude/skills/daily-workflow/SKILL.md:1)

• Purpose: Main skill file with frontmatter trigger description
• Auto-loaded by Claude Code when conversation matches triggers
• Contains: Quick workflow, classification rules, context areas, critical rules
• Token-efficient: ~350 lines (under 500 line best practice)

1. Extraction Script (.claude/skills/daily-workflow/scripts/extract_content.py:1)

• Purpose: Deterministic file I/O, pattern matching, data structure operations
• Key functions:
• find_daily_note(date_str) - Locate daily note files by date
• extract_notes_section(filepath) - Extract ### 🧠 Notes content
• split_entries(notes_content) - Split into individual entries
• check_duplicate(entry, existing_entries) - Fuzzy matching (80% threshold)
• format_cross_link(daily_note_path) - Generate wikilinks
• Handles: Tracker management, file creation with frontmatter, atomic appends

1. Reference Documentation (.claude/skills/daily-workflow/reference/)

• templates.md - Daily note structure, Dataview patterns, entry recognition
• formatting-rules.md - Transformation rules, cross-linking format, deduplication

Data Flow

Daily Note (00 - Daily/)
  ↓
### 🧠 Notes Section Extraction (Python script)
  ↓
Entry Splitting (by blank lines)
  ↓
AI Classification (Claude/MCP)
  ├─ Diary Entry (preserve narrative)
  ├─ Insight (preserve reflective tone)
  ├─ Context (transform to factual)
  └─ Idea (preserve speculative tone)
  ↓
Confidence Scoring (0-1.0)
  ↓
Routing Decision (>0.8 for secondary)
  ├─ Primary: Always route to main file
  └─ Secondary: Route to area-specific files (if >80% confidence)
  ↓
Transformation (for context entries only)
  ├─ Remove: pronouns, dates, feelings
  ├─ Extract: topical facts
  └─ Custom formulation: per context area
  ↓
Deduplication Check (fuzzy match 80%)
  ↓
Append to Destination Files
  ├─ With cross-links: *From:* [[daily-note#section]]
  └─ With related links: *Related:* [[context-area]]
  ↓
Smart Memory Detection
  ├─ Serena Memory (technical patterns)
  └─ Graphiti Memory (strategic insights)
  ↓
Tracker Update (mark processed, except today)

Integration Points

• Serena MCP: Pattern matching, file operations
• mcp__serena__list_dir - Find daily notes
• mcp__serena__search_for_pattern - Extract sections
• mcp__serena__find_file - Locate context files
• mcp__serena__write_memory - Update Serena memories

• Graphiti MCP: Knowledge graph storage (3 instances)
• graphiti-mokai - MOKAI business insights
• graphiti-finance - Financial insights
• graphiti-personal - Personal insights

• Daily Notes Template (98-templates/daily-note.md:327)
• ### 🧠 Notes section (extraction target)
• Dataview queries for tasks, events, projects

• Context Areas (24 files across PARA structure)
• Business: mokai, mokhouse, accounting, crypto, soletrader, smsf, safia, trust
• Personal Development: learning, mindset, psychedelics
• Health & Fitness: health-fitness, diet, medical, gym
• Tech: tech, ableton, mac, ai, engineering
• Other: claude-code, harry, people

Configuration

Skill Trigger Phrases

Auto-triggers when user says:

• "What did I work on today?"
• "Extract content from my daily note"
• "I'm done for today"
• "Wrapping up"
• "Daily summary"
• "End of day"
• References "### 🧠 Notes" section

Environment Variables

None required (uses Claude Code's built-in AI and MCP connections)

Setup Requirements

1. Enable skill in Claude Code:

• Settings > Capabilities > Skills > Enable
• Restart Claude Code after enabling

1. Ensure daily note template is configured:

• Template location: 98-templates/daily-note.md
• Must include ### 🧠 Notes section

1. Verify destination files exist:

• 04-resources/diary.md
• 04-resources/context.md
• 01-areas/p-dev/insights.md
• 04-resources/ideas.md

1. Optional: Pre-create area-specific files (auto-created if missing):

• business/mokai/diary-mokai.md
• business/mokai/CLAUDE.md
• etc.

Usage

Auto-Triggering (Primary Method)

Natural conversation triggers:

User: "What did I work on today?"
→ Skill activates, scans today's daily note, extracts and classifies entries
 
User: "Extract content from my daily note"
→ Skill activates with explicit extraction intent
 
User: "I'm done for today, extract my notes"
→ End-of-day trigger, processes today's entries

Manual Script Execution (Testing/Debugging)

# Extract specific date
python .claude/skills/daily-workflow/scripts/extract_content.py 2025-10-21
 
# Scan for unprocessed notes
python .claude/skills/daily-workflow/scripts/extract_content.py --scan-unprocessed

Common Scenarios

Scenario 1: End-of-Day Extraction

User: "Done for today"
Claude: [Activates daily-workflow skill]
  1. Scans today's daily note
  2. Extracts 5 entries from ### 🧠 Notes
  3. Classifies: 2 Diary, 1 Insight, 1 Context, 1 Idea
  4. Presents routing plan with confidence scores
  5. User approves: "y"
  6. Routes entries to appropriate files
  7. Detects memory opportunities (Serena + Graphiti)
  8. Updates tracker

Scenario 2: Specific Date Extraction

User: "Extract content from my note on October 19th"
Claude: [Activates daily-workflow skill with date context]
  1. Locates 🌤️ Fri - 19th Oct 25.md
  2. Extracts and processes entries
  3. Routes to destinations

Scenario 3: Batch Processing

User: "Process all unprocessed daily notes from this week"
Claude: [Activates daily-workflow skill]
  1. Scans for unprocessed notes (excluding today)
  2. Processes each chronologically
  3. Presents combined routing decisions
  4. Executes after approval

Approval Flow

After classification, user chooses:

• [y]es - Approve all routing decisions and execute
• [n]o - Cancel extraction
• [e]dit - Modify routing decisions before execution
• [s]kip entry - Skip specific entries

For memory updates:

• [a] Update all - Update all suggested memories (confidence >0.80)
• [s] Selective - Choose which memories to update
• [n] Skip - No memory updates
• [l] Later - Queue for end-of-day batch review

Examples

Example 1: Daily Diary Entry to Context Transformation

Original Daily Note Entry (00 - Daily/🌤️ Mon - 21st Oct 25.md):

### 🧠 Notes
 
Today I built a Supabase database for financial tracking for MOKAI. Took forever to set up auth but got it working. Will use for invoice management.

AI Classification:

• Type: Context (96% confidence)
• Reason: Factual technical information
• Primary: 04-resources/context.md
• Secondary: business/mokai/CLAUDE.md (98%), tech/CLAUDE.md (94%)

Routed to 04-resources/context.md:

-Supabase: Financial tracking database with auth, for invoice management
  *From:* [[00 - Daily/🌤️ Mon - 21st Oct 25#🧠 Notes]]

Routed to tech/CLAUDE.md:

-Supabase: Financial tracking database with auth
  *From:* [[00 - Daily/🌤️ Mon - 21st Oct 25#🧠 Notes]]
  *Related:* [[business/mokai/context-mokai]]

Routed to business/mokai/CLAUDE.md:

-Financial tracking system: Supabase database for invoice management
  *From:* [[00 - Daily/🌤️ Mon - 21st Oct 25#🧠 Notes]]
  *Related:* [[tech/context-tech]]

Example 2: Smart Memory Detection

Entry: "Created new workflow using Supabase triggers for invoice automation"

Detection:

System: Serena
Memory: system_patterns_and_guidelines
Confidence: 0.87
Update: "Invoice automation: Supabase triggers reduce manual data entry"
 
System: Graphiti
Instance: graphiti-mokai
Confidence: 0.85
Episode: "MOKAI invoice automation using Supabase triggers for financial workflow optimization"

Example 3: Multi-Area Routing

Entry: "Finished mixing Nintendo track for MOK HOUSE client. Client approved on first submission."

Classification:

• Type: Diary (94% confidence)
• Primary: 04-resources/diary.md
• Secondary: business/mokhouse/diary-mokhouse.md (97%)

Result:

• Original narrative preserved in both files
• Cross-links added bidirectionally
• No transformation (diary entries keep narrative voice)

Dependencies

External Services

None (uses Claude Code's built-in capabilities)

Python Packages (Built-in)

• os, re, json, sys - Standard library
• pathlib - File path operations
• datetime - Date parsing and formatting
• difflib - Fuzzy matching for deduplication

Internal Dependencies

• Claude Code Skills System - Auto-trigger mechanism
• Serena MCP - File operations, pattern matching, memory management
• Graphiti MCP - Knowledge graph storage (3 instances)
• Daily Notes Template - ### 🧠 Notes section format
• PARA Structure - Context area organization (01-areas, 04-resources)

Troubleshooting

Common Issues

Issue 1: Skill not triggering automatically

• Cause: Skills not enabled in Claude Code settings
• Solution: Settings > Capabilities > Skills > Enable, then restart Claude Code

Issue 2: Missing `### 🧠 Notes` section in extraction

• Cause: Daily note uses different heading format
• Solution: Ensure daily note template matches exact format (case-sensitive, emoji included)

Issue 3: Entries classified incorrectly

• Cause: Ambiguous entry content or insufficient context
• Solution: Use edit mode [e] to correct classification, AI learns from patterns

Issue 4: Too many secondary destinations

• Cause: Confidence threshold too low
• Solution: Currently set to 0.8 (80%), can be increased for stricter routing

Issue 5: Diary files not auto-created

• Cause: File permission issues or incorrect path
• Solution: Verify write permissions, check directory structure exists

Issue 6: Duplicate entries appended

• Cause: Fuzzy matching threshold too low
• Solution: Currently 80% similarity, can be increased to prevent duplicates

Issue 7: Memory detection missing opportunities

• Cause: Entry doesn't contain trigger keywords
• Solution: Review detection keywords in SKILL.md:119, adjust confidence threshold

Issue 8: Wrong Graphiti instance selected

• Cause: Routing logic based on area doesn't match entry content
• Solution: Use selective mode [s] to manually specify correct instance

Debugging

# Check tracker status
cat .claude/commands/.extract-daily-content-tracker.json | jq '.'
 
# View last processed file
cat .claude/commands/.extract-daily-content-tracker.json | jq '.processed_files | to_entries | last'
 
# Reset tracker (force re-extraction)
rm .claude/commands/.extract-daily-content-tracker.json
 
# Test Python script directly
python .claude/skills/daily-workflow/scripts/extract_content.py --scan-unprocessed
 
# Verify daily note format
grep -A 10 "### 🧠 Notes" "00 - Daily/🌤️ $(date +"%a - %eth %b %y").md"

Monitoring & Maintenance

Logs

• Skill activation: Visible in Claude Code conversation (no separate logs)
• Tracker file: .claude/commands/.extract-daily-content-tracker.json tracks:
• Last scan timestamp
• Processed files with modification times
• Entries processed per file
• Routing decisions with confidence scores

Monitoring

# Check recent extractions
jq '.processed_files | to_entries | map({date: .value.processed_at, count: .value.entries_count}) | sort_by(.date) | reverse | .[0:5]' .claude/commands/.extract-daily-content-tracker.json
 
# Total entries processed
jq '.entries_processed' .claude/commands/.extract-daily-content-tracker.json

Update Frequency

• Skill logic: Update SKILL.md when workflow changes
• Templates: Update reference/templates.md when daily note format changes
• Formatting rules: Update reference/formatting-rules.md when transformation rules change
• Python script: Update when new extraction features needed

Related Systems

• MOKAI Daily Operations Skill - MOKAI-specific daily workflow
• MOK HOUSE Operations Skill - MOK HOUSE business workflow
• Extract Daily Content Command - Original manual command (deprecated)
• Daily Note Template - Template with extraction target section
• Serena MCP Memory - Memory management system
• Graphiti Knowledge Graph - Knowledge graph storage

Future Enhancements

Phase 2 (Planned)

1. Auto-categorization during writing: Real-time classification as user types in daily note
2. Batch processing UI: Interactive dashboard for reviewing and approving batches
3. Machine learning improvements: Learn from user corrections to improve classification accuracy
4. Custom context areas: User-defined context areas beyond the 24 default ones
5. Export functionality: Export extracted data to external tools (Notion, Roam, etc.)

Phase 3 (Exploration)

1. Voice input integration: Extract from voice-dictated daily notes
2. Mobile app sync: Sync extractions from mobile daily notes
3. Collaboration features: Share context/insights with team members
4. Analytics dashboard: Visualize entry patterns, productivity trends
5. AI summarization: Generate weekly/monthly summaries from extracted entries

Change Log

• 2025-10-21: Initial implementation
• Created skill structure with SKILL.md, reference files, Python script
• Implemented auto-trigger mechanism based on conversation context
• Added AI classification for 4 entry types (Diary/Insight/Context/Idea)
• Configured routing to 24 context areas with confidence scoring
• Integrated Serena and Graphiti memory detection
• Documented comprehensive formatting and transformation rules
• Created deduplication logic with 80% fuzzy matching
• Implemented tracker system for incremental processing
• Added cross-linking and bidirectional references
• Updated skills README with testing instructions

Testing & Validation

Manual Testing Checklist

After implementing or modifying the skill:

• [ ] Skill triggers on phrase "What did I work on today?"
• [ ] Extracts entries from ### 🧠 Notes section correctly
• [ ] Classifies entries into correct types (Diary/Insight/Context/Idea)
• [ ] Routes to primary destinations (all types)
• [ ] Routes to secondary destinations when confidence >80%
• [ ] Transforms diary narratives to factual statements for context files
• [ ] Creates cross-links with correct format
• [ ] Auto-creates missing diary files with frontmatter
• [ ] Deduplicates entries (fuzzy match 80%)
• [ ] Updates tracker after successful extraction
• [ ] Does NOT mark today's note as processed
• [ ] Detects Serena memory opportunities
• [ ] Detects Graphiti memory opportunities (correct instance)
• [ ] Presents approval flow to user
• [ ] Handles user corrections in edit mode

Validation Commands

# Verify skill files exist
ls -lah .claude/skills/daily-workflow/
ls -lah .claude/skills/daily-workflow/reference/
ls -lah .claude/skills/daily-workflow/scripts/
 
# Check SKILL.md line count (should be <500)
wc -l .claude/skills/daily-workflow/SKILL.md
 
# Verify Python script is executable
ls -l .claude/skills/daily-workflow/scripts/extract_content.py | grep -q 'x' && echo "Executable" || echo "Not executable"
 
# Test Python script
python .claude/skills/daily-workflow/scripts/extract_content.py --scan-unprocessed

Success Criteria

• ✅ Correctly classify entry types (>90% accuracy with AI)
• ✅ Route to appropriate areas with good confidence (>80% for secondary)
• ✅ Show clear reasoning for routing decisions
• ✅ Allow easy manual override via approval flow
• ✅ Create valid cross-links bidirectionally
• ✅ Auto-create diary files with proper frontmatter
• ✅ Track all routing decisions in tracker JSON
• ✅ Handle incremental runs efficiently (skip already processed)
• ✅ Maintain file integrity across all destinations
• ✅ Respect >80% confidence threshold for secondary routing
• ✅ Transform diary to context correctly (remove pronouns, dates, feelings)
• ✅ Preserve narrative in diary files
• ✅ Detect memory opportunities with >75% confidence

Notes

• Token Efficiency: SKILL.md kept under 500 lines, uses progressive disclosure pattern
• Auto-Detection: Triggers automatically based on frontmatter description
• Hybrid Approach: Skill for common workflows, command as manual fallback
• Smart Context: Only loads reference files when needed (not pre-loaded)
• Incremental Processing: Tracker ensures files aren't re-processed unnecessarily
• Today's Note: Always re-processes today's note (allows same-day updates)
• Deduplication: 80% fuzzy match threshold prevents near-duplicates
• Confidence Threshold: 80% for secondary routing balances precision and recall
• Memory Detection: Suggests Serena/Graphiti updates but requires user approval
• Cross-Linking: Bidirectional links maintain vault connectivity