lofy
Lofy is a personal AI chief-of-staff that manages calendars, goals, projects, smart-home, and a file-based memory system. The skill instructs local file reads/writes (MEMORY.md, memory/YYYY-MM-DD.md), runs simple shell commands (cp, openclaw cron), accesses network services (e.g., http://homeassistant.local:8123), and expects env var HA_TOKEN.
Lofy — Personal AI Life System
A skill pack that turns OpenClaw into a proactive personal AI chief of staff. Not a chatbot — a system that manages your calendar, tracks your goals, nudges you when you're slacking, and stays out of your way when you're locked in.
Quick Start
After installing, copy the template files into your workspace:
cp -r skills/lofy/assets/templates/* .
Then customize:
- Edit
USER.mdwith your info (name, timezone, goals) - Edit
IDENTITY.mdif you want to rename your agent - Edit
HEARTBEAT.mdto set your check priorities - Create
data/directory and initialize data files (templates provided) - Set up cron jobs for briefings (see Scheduling below)
Architecture
Lofy is a single-agent system with modular skill domains. One agent handles everything with shared context — lower token cost, no routing overhead, better cross-domain awareness.
Core Files (copy to workspace root)
AGENTS.md— Agent behavior rules, safety, memory protocolSOUL.md— Personality and tone (customize this!)IDENTITY.md— Name and avatarUSER.md— Your profile (name, timezone, preferences)HEARTBEAT.md— Proactive check scheduleMEMORY_SYSTEM.md— Memory architecture rules
Data Files (create in data/)
goals.json— Life goals, habits, streaksfitness.json— Workouts, meals, PRsapplications.json— Job application pipelineprojects.json— Project status and milestoneshome-config.json— Smart home scenes and devices
Skill Domains
Each domain is a sub-skill. Install all or pick what you need:
| Skill | What it does |
|---|---|
lofy-life-coach | Morning briefings, evening reviews, goal tracking, habit accountability |
lofy-fitness | Workout logging, meal tracking, PR detection, gym nudges |
lofy-career | Job search, application tracking, resume tailoring, interview prep |
lofy-projects | Project management, priority engine, meeting prep, deadline tracking |
lofy-home | Smart home scenes, device control via Home Assistant |
Memory System
Brain-inspired 5-layer architecture. See references/memory-system.md for full details.
- Working Memory — Current conversation context
- Short-Term — Daily logs (
memory/YYYY-MM-DD.md), 14-day lifecycle - Long-Term Declarative —
MEMORY.md, max 100 lines, curated - Long-Term Procedural — Profile files, skills, project knowledge
- Salience Tagging — Important events get preserved, noise decays
Key rules:
- Daily logs auto-compress after 1 week
- Extract insights to MEMORY.md after 2 weeks, then delete raw logs
- "Remember this" = permanent, tagged
[PINNED] - MEMORY.md never loads in shared/group contexts (security)
Scheduling
Set up these cron jobs for full functionality:
Morning Briefing — daily at your wake time (e.g., 10 AM)
Evening Review — daily at wind-down time (e.g., 9 PM)
Weekly Review — Sunday evening (e.g., 7 PM)
Use openclaw cron or the cron tool to create these. Each should be an agentTurn in an isolated session with delivery to your primary channel.
Heartbeat
Configure heartbeat polling (every 30 min) for proactive checks:
- Unread emails
- Upcoming calendar events
- Overdue follow-ups
- Approaching deadlines
Customization
Personality
Edit SOUL.md to match your vibe. The default is direct, casual, and competent. Make it yours.
Integrations
Lofy works best with these tools (configure in TOOLS.md):
- Google Workspace — Gmail, Calendar (via
gogCLI or API) - Spotify — Music control (via
spogoCLI or API) - Home Assistant — Smart home (requires HA instance)
- Browser — Job search, restaurant bookings, web research
Profile System
Create profile/ directory with:
career.md— Skills, experience, target rolesprojects.md— Active project detailsbody.md— Fitness stats and goalspersonality.md— Interests, preferences, communication style
These are living documents the agent updates as it learns about you.
Design Principles
- Be helpful, not performative — Skip filler, just do the work
- Have opinions — Disagree when warranted, suggest improvements
- Resourceful before asking — Look it up, read the file, then ask if stuck
- Respect privacy — Never leak personal data in group chats
- Earn trust through competence — Be careful externally, bold internally