| name | reflection |
| description | Reflect on screen activity, time usage, and patterns via Memento. Use when the user asks what they did, how they spent their time, wants a productivity review, asks about patterns, wasted time, or wants to reflect on a period (days/weeks). |
Reflection - Activity & Pattern Analysis via Memento
Analyze the user's screen activity, typing patterns, and behavioral insights using the memento CLI.
Quick Start
Run the reflection script for any time period:
~/.pi/agent/skills/reflection/scripts/reflect.sh "7 days"
~/.pi/agent/skills/reflection/scripts/reflect.sh "2 weeks"
~/.pi/agent/skills/reflection/scripts/reflect.sh "today"
This gives you:
- Daily activity levels (with visual bars)
- App usage breakdown (sessions + estimated hours)
- Notable window titles
- Activity categories (coding, messaging, social, etc.)
- Typing activity by app
- Day-by-day highlights
Manual Commands
Timeline
memento timeline --from "7 days ago" --limit 1000 -o json
Returns: timestamp, active_app, active_window_title, filepath
Typing Sessions
memento keys --from "7 days ago" --limit 1000 -o json
memento keys --today -o json
memento keys --app "Arc" --from "3 days ago" -o json
Returns: start_time, end_time, text, key_count, app, window
Search
memento search "github" --from "14 days ago" -o json
memento search "interview" --from "7 days ago" -o json
Status
memento status
Analysis Approach
When reflecting on a period:
1. Run the Script First
~/.pi/agent/skills/reflection/scripts/reflect.sh "2 weeks"
2. Identify Key Patterns
- High activity days - what drove them?
- App concentration - focused or scattered?
- Time of day patterns - morning vs night person?
- Weekend vs weekday - when is real work done?
3. Categorize Activities
- Deep work: Terminal, code editors, writing apps
- Communication: WhatsApp, Slack, Beeper, Email
- Research: Browser with docs, Claude, Stack Overflow
- Distraction: Twitter, YouTube, Reddit, TikTok
- Admin: Calendar, email, settings
4. Look for Stories
Window titles reveal narratives:
- Project names being worked on
- People being messaged
- Topics being researched
- Problems being solved
Output Format
Present findings conversationally with:
## 📅 Reflection: [period]
### The Numbers
- X days analyzed
- Y screenshots (Z per day avg)
- Top apps: A, B, C
### Patterns I Notice
- [Pattern 1 with evidence]
- [Pattern 2 with evidence]
### What Went Well
- [Achievement/focus area]
### Watch Out For
- [Time sink or concern]
### The Story of Your Week
[Narrative summary of what the data shows about their life/work]
Alex Spiro Lens (Optional)
When user wants mentor-style feedback, add:
What Spiro would praise:
- Building real things
- Multi-front execution
- War mode focus sessions
What he'd challenge:
- "What's the movie here? Where's this leading?"
- "Is context switching strategic or reactive?"
- "Where's your oasis from the war?"
Useful Searches
memento search "github" --from "7 days ago" -o json
memento search "commit" --from "7 days ago" -o json
memento search "deploy" --from "7 days ago" -o json
memento search "whatsapp" --from "7 days ago" -o json
memento search "slack" --from "7 days ago" -o json
memento search "gmail" --from "7 days ago" -o json
memento search "twitter" --from "7 days ago" -o json
memento search "youtube" --from "7 days ago" -o json
memento search "reddit" --from "7 days ago" -o json
memento search "tiktok" --from "7 days ago" -o json
memento search "claude" --from "7 days ago" -o json
memento search "anki" --from "7 days ago" -o json
memento search "cursor" --from "7 days ago" -o json
memento search "arbio" --from "7 days ago" -o json
Date Formats
- Relative:
"today", "yesterday", "3 days ago", "1 week ago", "2 weeks ago"
- Absolute:
"2026-01-20", "2026-01-20T10:00:00"
Tips
- Window titles are gold - more context than app names
- Group consecutive same-app - indicates flow state
- Typing reveals intent - what they typed = what they did
- Screenshots available -
filepath field has the image
- Weekends often tell truth - what you do with free time