antigravity-skills-reference/skills/diary

📔 Unified Diary System (Agentic Context-Preserving Logger) v4.1

Unified Diary System is a fully automated, anti-pollution AI journaling and synchronization workflow designed specifically for multi-project developers and creators. By leveraging Continuous Tool Calling from AI Agents, a single natural language command automatically executes a 4-step pipeline: Local Project Logging ➔ Global Context Fusion ➔ Cloud Bi-directional Sync ➔ Experience Extraction, achieving a true "One-Shot" seamless record.

---

✨ Core Features

• ⚡ Agent One-Shot Execution: Once triggered, the AI completes the entire technical process without interruption, only pausing at the final step to ask for human validation on extracted "lessons learned".
• 🛡️ Context Firewall: Strictly separates "Project Local Diaries" from the "Global Master Diary." This fundamentally solves the severe "Context Pollution / Tag Drift" problem where AI hallucinates and mixes up progress between Project A and Project B during daily summaries.
• 🧠 Automated Lessons Learned: More than just a timeline of events, the AI proactively extracts "New Rules" or "Optimizations" from the bugs you faced or discoveries you made today, distilling them into your Knowledge Base.
• 🔄 Seamless Cross-Platform Sync: Includes built-in scripts to push the final global diary straight to Notion and/or Obsidian with a simple --sync-only flag.

---

🏗️ The 5-Step Workflow Architecture

When a developer types :{Write a diary entry using the diary skill} in any project directory, the system strictly executes the following atomic operations:

Step 1: Local Project Archiving (AI Execution)

1. Auto-Location: The AI calls terminal commands (e.g., pwd) to identify the current working directory, establishing the "Project Name".
2. Precision Writing: It writes today's Git Commits, code changes, and problem solutions in "append mode" exclusively into that project's local directory: diary/YYYY/MM/YYYY-MM-DD-<Project_Name>.md.

Step 1.5: Refresh Project Context (Automation Script)

• Auto-Execution: The AI invokes prepare_context.py to scan the project's latest directory structure, tech stack, and diary-based action items, generating/updating the AGENT_CONTEXT.md at the project root.

Step 2: Extracting Global & Project Material (Automation Script)

• Material Fetching: The AI automatically executes fetch_diaries.py, precisely pulling the "just-written local project diary" and today's "Global Diary (if it exists)", printing both to the terminal for the AI to read.

Step 3: AI Smart Fusion & Global Archiving (AI Execution)

• Seamless Fusion: The AI mentally sews the two sources from Step 2 together, writing the combined result into the global diary vault: .../global_skills/auto-skill/diary/YYYY/MM/YYYY-MM-DD.md.
• Strict Zoning: It uses ### 📁 <Project Name> tagging to ensure existing project progress is preserved, while new project progress is safely appended—absolutely no overwriting.

Step 4: Cloud Sync & Experience Extraction (Script + Human)

1. One-Click Push: The AI calls master_diary_sync.py --sync-only to push the data to Notion/Obsidian.
2. Human Authorization: The AI extracts today's 📌 New Rules or 🔄 Experience Optimizations and presents them to the developer. Once authorized, these are written to the local Knowledge Base and embedded (e.g., via qmd embed).

---

📂 Directory Structure

This system adopts a "Distributed Recording, Centralized Management" architecture:

📦 Your Computer Environment
 ┣ 📂 Project A (e.g., auto-video-editor)
 ┃ ┗ 📂 diary/YYYY/MM/
 ┃    ┗ 📜 2026-02-24-auto-video-editor.md  <-- Step 1 writes here (Clean, isolated history)
 ┣ 📂 Project B (e.g., GSS)
 ┃ ┗ 📂 diary/YYYY/MM/
 ┃    ┗ 📜 2026-02-24-GSS.md                
 ┃
 ┗ 📂 Global Skills & Diary Center (This Repo)
    ┣ 📂 scripts/
    ┃  ┣ 📜 fetch_diaries.py                <-- Step 2: Material transporter
    ┃  ┣ 📜 prepare_context.py              <-- Step 1.5: Context refresher
    ┃  ┗ 📜 master_diary_sync.py            <-- Step 4: Notion/Obsidian sync
    ┣ 📂 knowledge-base/                    <-- Step 4: AI extracted lessons
    ┗ 📂 diary/YYYY/MM/
       ┗ 📜 2026-02-24.md                   <-- Step 3: The ultimate fused global log

---

🚀 How to Use (Usage)

After setting up .env with your Notion tokens, simply input the following into your CLI/IDE chat while working inside a project:

:{Write a diary entry using the diary skill} Today I finished the initial integration of the Google Colab python script and fixed the package version conflicts.

The system will take over to handle all the filing, merging, and syncing automatically.

---

🛠️ Setup & Prerequisites

1. Configuration: Rename .env.example to .env and fill in your NOTION_TOKEN, NOTION_DIARY_DB, and set where your global diary root is stored.
2. Dependencies: pip install -r requirements.txt
3. AI Agent: Requires an AI assistant with Function Calling / Continuous Tool Calling capabilities (like Cursor, Claude Code, or Gemini CLI frameworks).

---

💡 Design Philosophy: Why not just have the AI write directly to the global diary? Because we found that when an AI lacks the "isolated local project context", it frequently suffers from Tag Drift (writing Project A's progress under Project B's header). Through this highly-structured "Local First, Global Second" 4-step architecture, we completely eliminated the context pollution pain point in AI-automated logging.