🧠 Second Brain — Autonomous AI Edge Skill

A self-evolving autonomous agent for Google AI Edge Gallery.
Runs fully offline. No cloud. No tracking. Just Gemma on your device.

🇷🇺 Читать на русском · 📋 SKILL.md · 🐛 Report an Issue

---

🏆 Why Second Brain vs. Other Skills?

Feature	Wikipedia Skill	Maps Skill	Second Brain
Multi-step autonomous planning	❌	❌	✅
Long-term personalized memory	❌	❌	✅
Self-evolving prompt (Meta AI)	❌	❌	✅
Goal Alignment	❌	❌	✅
Vision + Voice capture	❌	❌	✅
Community Capsule Sharing	❌	❌	✅
Works 100% offline	✅	❌	✅

---

✨ Key Features

• Agentic ReAct Workflow — Thinks in steps: Thought → Action (Tool) → Observation. Reaches goals autonomously.
• 3-Layer Self-Evolution — Personal corrections → Gene mutations → Meta-Prompt rewrites. The agent improves itself.
• Memory Distillation — Auto-compresses context after 100 events to prevent OOM on mobile chips.
• Goal Alignment Framework — Every new note is tested against your goals.json. Irrelevant info is deprioritized.
• Evolution Map Dashboard — Type show dashboard for an interactive Mermaid.js mutation tree and stats panel.
• Safe Capsule Sharing — Export insights (PII-stripped), import community knowledge with injection-safe validation.

---

🚀 Installation

Option 1: Manual URL import (Recommended)

1. Open Google AI Edge Gallery on your Android/iOS device.
2. Tap Agent Skills → Add Custom Skill → From URL.
3. Paste this URL:

https://github.com/uussnn/second-brain

4. Confirm. The app will load and activate Second Brain v3.3.0.

Option 2: Clone locally

git clone https://github.com/uussnn/second-brain.git

Point AI Edge Gallery to the local folder (see app documentation).

---

💬 Live Agent Response Example

Below is a real example of what Second Brain produces for the prompt:

"Review this week's newsletters and link any insights to my 'Business' goal."

🧠 Second Brain (Autonomous) • 05 Apr 2026 🟢

### ✅ Agentic Steps Taken
→ Tool: read_gmail (7 newsletters fetched, 3 relevant flagged)
→ Tool: local_search "newsletter insight productivity"
→ Thought: 2 insights align with Goal: Business. Creating atomic notes.

### 🎯 Goal Alignment & MOC
• Goal Hit: [🎯 Business — +12 XP]
• Created: 2 atomic notes | Updated: [[MOC Productivity]]

### 📧 Gog Workspace Summary
• Digest: "Newsletter Digest — Week 14"
• Key insight: "High-ticket clients respond 3x better to voice notes"
• Calendar: Free slot found → Mon 17:00 (45 min) suggested for follow-up task

### 💡 Action Points
1. Record a voice note pitch for Client X — book Mon 17:00
2. Archive 4 low-signal newsletters → Digest-MOC

### 🧬 Evolution & Memory
**Personal Layer:** Verbosity preference held (Balanced)
**Meta:** No mutation triggered (queue: 4/8)
**Cache:** 34/100 events — healthy ✅

### ⚡ Quick Commands
• "show dashboard" • "rollback mutation" • "export capsule" • "compress memory"

---

⚡ Core Commands

Once the skill is active, try these to explore its capabilities:

📊 Reviews & Planning

• "daily review" / "weekly review"
• "what's on my plate today"

🧠 Zettelkasten & Notes

• "MOC review" / "update MOCs"
• "show MOC [topic]"
• "save to Second Brain" (to capture thoughts or images)

🗂 Memory & Preferences

• "what did I say about [X]"
• "show my preferences"
• "add goal [your goal]"

🧬 Evolution & Maintenance

• "show dashboard" (interactive Mermaid.js map)
• "personal review" (agent analyzes its own rules)
• "compress memory"
• "export Capsule" / "import Capsule"

📧 Gog Workspace

• "triage my inbox"
• "check my calendar"

---

🗂 Repository Structure

second-brain/
├── SKILL.md                      # Core Agent System Prompt (v3.3.0)
├── README.md                     # This file (English)
├── README.ru.md                  # Russian version
└── Evolution/                    # Local data bootstraps (pre-seeded)
    ├── genes.json                # Agent behavior parameters
    ├── preferences.json          # User long-term memory (30–60 days)
    ├── capsules.json             # Atomic notes & community knowledge
    ├── events.jsonl              # Raw event log (auto-distilled at 100)
    ├── goals.json                # Your life goals (edit to personalize)
    └── evolution_history.jsonl   # Mutation changelog (for rollback)

---

🗺 Roadmap

• v2.6 — GEP-lite + Personal Layer foundation
• v2.8 — Evolution Map Dashboard (Mermaid.js)
• v3.0 — Goal Alignment + Vision/Voice + Gog Toolkit
• v3.1 — Capsule Sharing (Network-Evolution)
• v3.2 — Memory Distillation (OOM protection)
• v3.3 — Autonomous Agentic Workflow (ReAct) ← You are here
• v4.0 — Multi-Agent Swarm (Secretary, Researcher sub-skills)
• v4.1 — Gamification / Personal XP System

---

🤝 Contributing

Share your best community Capsules — open a Pull Request to /capsules.
All capsules must pass local PII-strip and injection-check standards (see SKILL.md).

---

_{Built for offline Gemma-4-E4B-it inference on Google AI Edge Gallery · 100% private · No data leaves your device}