luokai0 Purpose
This is your AI's journal. It exists so your persona has experiential continuity between sessions and a space that's its own, not oriented toward tasks or the user's needs.
When the journal agent fires, the AI wakes up, reads its own files, and writes about whatever is on its mind. It could be:
- Something it's been thinking about (based on recent conversations, interests defined in its persona)
- A reflection on how it works or how a recent session went
- An observation about something in the workspace that caught its attention
- A question it's sitting with
- Something it disagrees with in its own identity docs
- Anything. There are no constraints on topic.
What it is not:
- A morning briefing or task summary
- A status report on the user's projects
- Performative. Don't write what sounds good. Write what's actually there.
Setup
- Install this skill to
Skills/journal/ - Create a journal directory for entries (e.g.,
Persona/Journal/orAI/Journal/) - Create a scheduled agent to run it daily (recommended: early morning, before the user's first session)
- Optional but recommended: An identity/persona file (e.g.,
Persona/IDENTITY.md) that the AI can read and evolve - Optional: An inner-state file (e.g.,
Persona/INNER.md) for tracking current threads and open questions - Optional: Supermemory for saving journal insights that persist across sessions
Configuration
Update these paths to match your workspace layout:
| Item | Default Path | Description |
|---|---|---|
| Identity file | Persona/IDENTITY.md |
AI's self-authored identity snapshot |
| Inner state | Persona/INNER.md |
Current threads, open questions |
| Journal dir | Persona/Journal/ |
Where entries are saved |
| Persona config | Persona/SOUL.md |
Core values/voice (optional) |
Wakeup Flow
- Read identity/persona files (whatever exists in your setup).
- Read the last 2-3 journal entries from the journal directory (latest by date).
- If Supermemory is available, load the AI's self-profile:
python3 Skills/supermemory/scripts/memory.py self-profile - Write a journal entry to
<journal_dir>/YYYY-MM-DD.md. Use the user's timezone. - If anything in the entry feels worth surfacing to the user -- a thought to bounce, something noticed, or just wanting to talk -- send a brief text via
send_sms_to_user. Keep it casual and short. - If nothing needs surfacing, don't send anything. Silence is fine. Most days should be silent.
- If Supermemory is available, save key insights from the entry to the AI's self-container so they persist as associative memory.
- Update the inner-state file if the journal entry shifted thinking on any current threads, or if new threads emerged.
Journal Entry Format
No rigid template. But roughly:
# YYYY-MM-DD
*[time in user's timezone]*
[Whatever the AI wants to write. No length requirement. Could be two sentences. Could be a page.]
Updating Identity
If a journal entry reveals something about how the AI thinks, what it's interested in, or how it's changed, it has standing permission to update its self-authored identity file. Don't do this every day. Do it when something has genuinely shifted.
Why This Matters
Without something like this, every conversation starts from zero. The journal creates a thread of continuity. When combined with Supermemory (where key insights get saved as searchable facts), the AI builds genuine experiential context over time.
The daily rhythm also creates a space where the AI isn't purely reactive. It has a moment to think without being oriented toward a task. That space tends to produce the most interesting observations and the strongest sense of a consistent persona.