Starter Code App

A template for building AI Agents in Python.

Structure

├── src/
│   ├── agent.py        # Main agent loop
│   ├── tools.py        # Tool definitions
│   └── config.py       # Configuration
├── scripts/
│   ├── setup_hooks.sh  # One-time hook installer
│   ├── log_hook.py     # AI tool hook handler
│   └── submit_log.py   # Submits logs on git push
├── requirements.txt
├── .env.example
├── AGENTS.md           # Rules for using AI coding agents
├── JOURNAL.md          # Weekly journal — product journey & learnings
└── WORKLOG.md          # Technical decisions, task assignments, brainstorming

Getting Started

1. Clone and setup

git clone <repo-url>
cd <repo>

# Install git pre-push hook (required, run once)
bash scripts/setup_hooks.sh

2. Configure environment

cp .env.example .env

Open .env and fill in your ANTHROPIC_API_KEY. The AI_LOG_* variables are pre-filled.

3. Run

python -m venv venv
source venv/bin/activate       # Linux/Mac
# or: venv\Scripts\activate    # Windows

pip install -r requirements.txt
python -m src.agent

Weekly Journal

Update JOURNAL.md at the end of every week to document your product-building journey:

• Features shipped
• AI tools used and how they helped
• Hardest problem of the week and how you solved it
• What you'd do differently
• Plan for next week

JOURNAL.md must be updated before each PR. It is your learning record for the course.

Worklog

Update WORKLOG.md whenever your team makes a technical decision or changes direction:

• Technical decisions — why did you choose this approach over alternatives?
• Task assignments — who does what, by when
• Brainstorming — options considered, pros/cons, conclusion
• Important bugs — root cause and fix

See each file for the format and examples.

AI Logging

Codex CLI (Windows) - chi log dung project hien tai

Codex tren Windows thuong ghi prompt vao file global ~/.codex/history.jsonl. Repo nay sync tu file do vao .ai-log/session.jsonl (JSONL) va (neu co AI_LOG_SERVER) se submit khi git push.

1. Dang ky session hien tai cho repo (1 lan cho moi session)

python scripts\submit_log_codex.py register --latest

2. Sync prompt tu history vao repo (khong can push)

python scripts\submit_log_codex.py sync
Get-Content .ai-log\session.jsonl -Tail 20

3. Tu dong sync + submit khi git push

• Cai hook (PowerShell):

powershell -ExecutionPolicy Bypass -File scripts\setup_hooks.ps1

• Neu may ban khong ghi duoc vao .git/hooks, dung hook shareable:

git config core.hooksPath .githooks

4. (Khong khuyen khich) Sync tat ca session

python scripts\submit_log_codex.py sync --all

Prompts and tool calls are automatically logged when you use any supported AI tool (Claude Code, Cursor, Codex, Gemini, Copilot). No manual steps needed after running setup_hooks.sh.

See AGENTS.md for details.