Inbox Processing Example

NOTE: This is an example skill showing how to build an inbox processing workflow. The actual inbox-processing skill is gitignored as it contains personal workflow patterns. Copy this to .claude/skills/inbox-processing/ and customize for your needs.

Overview

Process large Things3 inboxes (100+ items) efficiently through batch analysis, confidence-based automation, and intelligent user interaction.

Prerequisites:

•Load things3-productivity skill for MCP tool patterns
•Configure private-prefs/personal-taxonomy.json with your work areas and tags
•Create temp/inbox-processing/ folder for session state

When to use: Inbox has 100+ items requiring organization.

Personal Organization Integration

The skill uses LLM-driven analysis with context from personal-taxonomy.json:

•Work identification: Your configured work tags and areas
•Priority system: Your 1-9 priority scale
•Project patterns: Your existing projects and content patterns
•Semantic matching: Based on meaning, not just keywords

Core Workflow: Batch Processing

Phase 1: Initialize & Analyze

Step 1: Setup Session Create temp/inbox-processing/ with tracking files:

code

session.md                    # Batch progress, statistics
match_results.json            # Decisions with confidence scores
pending_decisions.json        # Items awaiting approval
high_confidence_actions.json  # Auto-apply candidates (≥90%)
execution_log.md              # Complete action history

Step 2: Load Inbox Batch

python

# First batch: 50 items, subsequent: 50-100 items
read_tasks(when="inbox", limit=50, include_notes=True)

Step 3: Load System Inventory Cache once per session:

python

list_areas()    # All areas with IDs and tags
list_projects() # All projects with metadata
list_tags()     # All tags including hierarchy

Phase 2: Confidence-Based Analysis

Step 4: Analyze Each Item For each inbox item, determine:

•Area assignment (90%+ confidence threshold)
•Project assignment (85%+ confidence threshold)
•Tag additions (based on content and context)
•Reference detection (notes without actionable tasks)

Confidence Levels:

•90-100%: Auto-apply safe (e.g., "Work: Fix bug" → area="Work")
•80-89%: Present for batch approval
•Below 80%: Skip, handle manually

Phase 3: User Interaction

Step 5: Present High-Confidence Batch Group by action type:

markdown

### Area: Work (25 items, 90-100% confidence)

Auto-assign these 25 items to area="Work"?
- "Work: Fix login bug" (100%)
- "Dashboard review" (95%)
...

[Approve] [Review individually] [Skip]

Step 6: Handle Medium-Confidence Items Present individually for 80-89% confidence:

markdown

1. "Design review notes" (85%) → area="Work"?
   Notes: Contains work-related keywords
   [Yes] [No] [Different area]

Phase 4: Execution

Step 7: Execute Approved Actions Batch operations by type:

python

# Set areas
edit_task(task_uuid="...", area="Work")

# Add tags
add_tags(task_uuids=[...], tags=["urgent"])

# Create projects
create_project(name="Q4 Roadmap", area="Work")

Step 8: Handle Reference Items Items with notes but no actionable task:

python

# Suggest migration to Notion
migrate_inbox_to_notion(block_id="your-block-id")

Phase 5: Completion

Step 9: Update Statistics Track in session.md:

code

Batch 1: 50 items processed
- 25 auto-assigned to areas
- 10 tagged
- 5 moved to projects
- 10 pending review

Remaining: 96 items

Step 10: Next Batch If inbox > 0, repeat from Step 2 with larger batch size (up to 100).

Pattern Learning

The skill improves through use:

•Successful matches reinforce confidence thresholds
•User corrections inform future suggestions
•Project creation patterns learned from history
•Tag combinations tracked for consistency

Example Confidence Scoring

markdown

### High Confidence (90-100%)

"Work: Fix dashboard bug" → area="Work"
- Explicit area mention (100%)
- Work tag keyword present
- Matches existing area pattern

### Medium Confidence (80-89%)

"Review team notes" → area="Work"?
- Work context implied (85%)
- No explicit area mention
- Could be personal or work

### Low Confidence (<80%)

"Call mom" → ???
- No clear work/personal indicators
- No matching patterns
- Requires manual classification

Customization Guide

To create your own inbox processing skill:

•

Copy this example:

bash

cp -r .claude/skills/inbox-processing.example .claude/skills/inbox-processing

•
Update references:
- •Replace "Work" with your actual work area names
- •Add your specific project patterns
- •Customize confidence thresholds

•

Configure personal-taxonomy.json:

json

{
  "things3": {
    "work_classification": {
      "work_tag": "YOUR_WORK_TAG",
      "work_areas": ["Your Work Area"]
    }
  }
}

•
Test with small batches:
- •Start with 10-20 items
- •Adjust confidence thresholds
- •Build pattern database

Tips

•Start conservative: Use higher confidence thresholds (95%+) initially
•Batch approvals: Group similar actions for efficiency
•Reference items: Migrate notes to Notion early to reduce inbox clutter
•Project discovery: Use list_projects=True to avoid creating duplicates
•Session breaks: Process in 30-minute focused sessions

Integration with Other Skills

•things3-productivity: Tool usage patterns and query strategies
•notion-workflows: Reference item migration patterns
•productivity-integration: Cross-system automation