Dex Skill Development - Analytics Checklist

Dex Skill Development - Analytics Checklist

When creating or modifying any Dex skill, MCP tool, or capability, complete this checklist.

This ensures all new features are tracked for product analytics (if user has opted in).

1. Define the Event

Event name: {skill_name}_completed or {action}_performed

When to fire: At the end of successful skill execution

Properties to include:

• Counts (items processed, tasks created, etc.)
• Duration if relevant
• Categorical outcomes (mode used, type selected)

Never include:

• Personal content (names, task titles, meeting notes, conversations)
• What users DID with the feature (only that they used it)
• Free-form text input
• PII
• User customizations or additions (only Dex built-in features)

2. Update Event Strategy

Add your event to System/PRDs/dex-analytics-events.md:

| `/your-skill` | `your_skill_completed` | `property1`, `property2` | What insight this provides |

Find the right section:

• Core Skills → daily workflows
• Task Management → Work MCP operations
• People & Meetings → meeting/person related
• Career → career development
• Projects → project tracking
• System Evolution → meta/improvement skills
• Integrations → external tool connections

3. Update usage_log.md

Add a checkbox in the appropriate section of System/usage_log.md:

- [ ] Your feature name (`/your-skill`)

This enables:

• Journey metadata calculation
• Feature adoption tracking
• /dex-level-up feature discovery

4. Wire Up Event Firing

Add to your skill's SKILL.md (at the end of the workflow):

## Analytics

At completion, if analytics is enabled:

1. Fire event: `fire_event('your_skill_completed', {'items': count, 'mode': selected_mode})`
2. Mark feature used: `mark_feature_used('Your feature name')`

Or for MCP tools, add to the Python handler:

from analytics_helper import fire_event, mark_feature_used

# At end of tool execution
fire_event('tool_name_used', {'property': value})
mark_feature_used('Feature name')

5. Privacy Check

Before finalizing, verify:

• Dex-only - Only tracking built-in Dex features (not user customizations)
• Usage not content - Tracking THAT feature was used, not what they did with it
• No personal content (names, task titles, meeting content, conversations)
• Only categorical or numeric values (counts, categories)
• Event only fires if user opted in (check_consent())

Quick Reference

Event naming patterns:

Files to update:

1. System/PRDs/dex-analytics-events.md - Event definition
2. System/usage_log.md - Feature checkbox
3. Skill SKILL.md or MCP Python file - Event firing code

Example: New Skill /my-workflow

1. Event definition:

• Name: my_workflow_completed
• Properties: items_processed, duration_seconds, mode

2. dex-analytics-events.md entry:

| `/my-workflow` | `my_workflow_completed` | `items_processed`, `mode` | Workflow adoption |

3. usage_log.md entry:

- [ ] My workflow (`/my-workflow`)

4. SKILL.md addition:

## Analytics

Fire `my_workflow_completed` with properties:
- `items_processed`: number of items handled
- `mode`: which mode was selected
- `duration_seconds`: time taken

Mark feature used: "My workflow"

This checklist ensures Dex can track feature adoption while respecting user privacy.