Codex Gigas: A Digital Garden as a Base for Agent Skills

Codex Gigas is the digital garden behind this study plan: every note you read here also feeds the skills described in agentic-best-practices.

Two roles, one repository

The project started as a study notebook but grew into two distinct, complementary roles:

Study material base: atomic notes in src/notas/, connected through [[wikilinks]] and grouped by category (Frontend, Backend, LLM, Computer Vision, and so on). The goal is retaining and reviewing technical knowledge — from React to object-detection algorithms like r-cnn.
Skill factory for AI agents: the same content becomes input for .claude/skills/, folders that package a workflow (e.g. skill-ui-validation) into reusable instructions any other project can pick up.

How a note becomes a skill

The path from idea to automation always follows the same direction:

Note first. A recurring problem (e.g. CSS breaking without the agent noticing) becomes a note documenting the “why” and the “how”.
Procedure extraction. If the workflow is repeatable and worth locking down, it migrates into a SKILL.md — a rigid contract of steps any agent must follow.
Routing via CLAUDE.md. The project’s master file points the trigger (“whenever CSS changes…”) to the skill, without cluttering global instructions with the full procedure.
Portability. Since a skill is just Markdown plus scripts, it’s copyable into .claude/skills/ of any other repository — Codex Gigas works as a catalog of skills already validated in production inside this very garden.

⚠️ Pitfall: not every note becomes a skill. Extraction only pays off when the process is repetitive, has verifiable steps (lint, tests, screenshots), and saves context that would otherwise need re-explaining every session.

Why bilingual

Every note is born as a pair (<slug>.md PT-BR + <slug>.en.md EN) because part of the goal is practicing technical-english — one of the gaps identified in the study plan — while the technical content stays useful for both agents and people, regardless of the source language.