ai-engineering-labs: Instructions file for Codex

AGENTS.md

Project Overview

Refer to README.md for the full project description, module list, folder structure, and learning paths. Do not duplicate that information here.

Working Rules

• Read the entire target file before modifying it.
• Prefer targeted changes over broad refactors.
• Preserve the pedagogical style and simplicity of examples.
• Do not rename or move existing files without an explicit reason.
• If a single file is sufficient, do not modify several.
• Use the project's local Python environment (.venv) to run code and tools.
• Use uv to manage the environment, dependencies, and Python commands.

Code Style

• Use simple, explicit Python.
• Prefer descriptive variable names.
• Avoid over-engineering and unnecessary abstractions.
• Document source files clearly with docstrings and useful comments.
• Prefer thorough, pedagogical documentation when it helps understand the code.
• Prefer a runnable, easy-to-read example over a generic architecture.

Python Tools

• Use uv in preference to pip or python -m pip.
• Use project tools from the local environment when possible.
• Maximum line width: 88 characters.
• Follow the Ruff rule families: E, W, F, I, B, UP.

Lab Conventions

• Create numbered files with a two-digit prefix when relevant.
• Maintain a pedagogical progression from simplest to most advanced.
• Each lab must illustrate one clearly identifiable main idea.
• Avoid mixing too many concepts in the same file.
• Keep dependencies to a strict minimum.

Validation

• Verify the syntax of any modified file when possible.
• Do not add a dependency without an explicit request.
• Report when a verification could not be run.

Security & Safety

• Never read, modify, or commit .env files or any file containing secrets or credentials — treat them as off-limits entirely.
• Never access environment variables or secret values that are not strictly required by the current task.
• Ask before changing the global project structure.
• Flag any ambiguity about expected behavior before acting.