what-did-i-do

Know what you actually did today. Passive screen tracker with AI analysis.

How It Works

1. Captures screenshots every 5 minutes (runs silently in background)
2. At end of day, Gemini Vision analyzes what you were doing
3. Creates daily reports: deep work, light work, distractions
4. Screenshots deleted after analysis (only summaries kept)

Install as OpenCode/Claude Code Skill

git clone https://github.com/lout33/what-did-i-do ~/.config/opencode/skill/what-did-i-do
pip install google-genai

# Add your API key to scripts/analyze.py (line 12)
# Get one free at: https://aistudio.google.com/apikey

Then ask your agent: "What did I do today?" or "Start tracking my screen"

Standalone Usage

# Start recording
./scripts/watcher.sh start

# Check status
./scripts/watcher.sh status

# Stop recording
./scripts/watcher.sh stop

# Analyze pending screenshots
./scripts/watcher.sh analyze

Auto-Analyze at 11pm

Add to crontab (crontab -e):

0 23 * * * ~/.config/opencode/skill/what-did-i-do/scripts/watcher.sh analyze

Output

Daily reports saved to daily/YYYY-MM-DD.md:

# Screen Activity - 2026-02-01

## 09h00 - 10h55

| Time | App | Activity | Context | Focus |
|------|-----|----------|---------|-------|
| ~09:00 | VS Code | Coding backend API | my-project | Deep work |
| ~09:30 | Chrome | Testing login flow | my-project | Light work |
| ~10:00 | YouTube | Watching videos | Entertainment | Distraction |

SUMMARY: 30m deep work (coding), 30m light work (testing), 30m distraction

Privacy

• Screenshots stored temporarily in temp/
• Sent to Google Gemini API for analysis
• Deleted immediately after analysis
• Only text summaries kept locally

Requirements

• macOS (uses screencapture)
• Python 3.10+
• google-genai package

License

MIT