totvs/engpro-advpl-tlpp-skills

EngPro AI Agent Skills is a repository maintained by TOTVS, a Brazilian enterprise software company, providing structured instructions for AI coding agents that work with ADVPL and TLPP. ADVPL is the programming language used to customize TOTVS Protheus, an ERP system widely used in Brazil and Latin America for finance, HR, and business operations. TLPP is a newer companion language in the same ecosystem. The README is written primarily in Portuguese. In this project, a skill is a folder containing a SKILL.md file with structured guidance that tells an AI coding agent how to perform a specific development task consistently and correctly. The README describes a three-level design: a small YAML header that is always loaded in the agent's context, a full instruction body that loads when the agent determines the skill is relevant, and supplementary files such as templates or reference examples that the agent reads only as needed. This approach is intended to keep token usage low while still providing detailed guidance at the moment it matters. The repository includes 20 skills for ADVPL/TLPP development covering tasks such as generating MVC screens, creating REST API endpoints, writing ProtheusDOC-style code documentation, reviewing code for known quality problems, migrating legacy code, and optimizing SQL queries. A second collection of 14 general-purpose skills called Superpowers provides agent behaviors applicable across any project. Skills can be installed at the user level to apply to all projects, or at the project level to apply only to a specific codebase. Supported AI agents include Claude Code, Cursor, GitHub Copilot, Cline, Windsurf, and more than a dozen others. The repository also includes CLAUDE.md and AGENTS.md files containing background knowledge about the Protheus ecosystem that agents read before starting work. TOTVS states in the README that human review of AI-generated code remains necessary, and that AI systems can produce errors.