Developer experience

Agala AI

A portable operating system for AI-assisted product delivery, expressed as agents, focused skills, and stack-specific engineering contracts.

Role
Designer & author
Stack
Agent workflows · Markdown contracts · Nuxt · Go · Vue
View repository

The challenge

General-purpose coding agents can produce fast output while missing the local standards, role boundaries, and verification habits that make work dependable.

The approach

Decisions that shaped the work.

  1. 01

    Split broad delivery work into product, architecture, frontend, backend, QA, review, and repair responsibilities.

  2. 02

    Encoded repeatable tasks as focused skills for Nuxt, Vue repositories, Go endpoints, design prototypes, review, and Git workflow.

  3. 03

    Added stack contracts for Nuxt and Go-plus-Vue applications so agents start from a shared technical baseline.

  4. 04

    Kept the system readable and portable as plain files rather than binding it to one proprietary runtime.

Better context beats louder prompting

Agala AI is an experiment in turning engineering judgment into reusable operating context. Instead of asking one agent to improvise every role, it gives planning, design, implementation, and verification their own boundaries.

The repository is intentionally plain. Agents and skills are reviewable text; stack definitions make architectural expectations visible; narrowly-scoped recipes make recurring implementation work consistent.

A practical orchestration model

The system covers the full delivery loop without pretending every task needs a large process. Small skills handle concrete work. Specialist agents own broader review or implementation responsibility. Global rules keep security and code quality expectations constant across both.

What it demonstrates

  • Product thinking applied to developer tooling
  • Agent orchestration through explicit responsibility boundaries
  • Reusable technical standards for Nuxt, Vue, and Go
  • A bias toward inspectable, vendor-light workflows

Built with intent in Buenos Aires.

© 2026 Julián Benitez