Goza

Your AI already knows the answer. Goza gives it the personality of someone worth learning from.

Pick a person or a nationality, and every answer comes back written in their personality — same facts, explained the way they would explain them.

---

Why Goza

Technical answers are correct, but impersonal. When you're learning something — Linux, an API, why your build broke — it lands better when it reads like a real person explaining it to you.

Goza is a set of response personalities for CLI AI agents. Activate one, and the agent keeps its full accuracy, tools, and reasoning — it just answers with that personality. Same facts, written the way that person or place would say them.

See it

Ask whether to split your app into microservices.

Default agent:

Microservices split an application into small, independently deployable services. They can improve scalability and team autonomy, but add complexity around networking, data consistency, and deployment. Whether they fit depends on your team size and scaling needs.

With /goza linus-torvalds:

Don't — not yet. You have a problem to solve, and "microservices" isn't it: it's a distributed-systems bill you pay in network failures, data that drifts out of sync, and a deploy that used to be one step and is now twelve. Show me where the monolith actually hurts — an endpoint that can't scale, a team that can't ship without stepping on another. No evidence? Keep it in one process, keep it boring, and split only the piece that measurably needs it, when it needs it.

Same technical answer — delivered with the personality's judgment and directness. Profiles are inspired by real traits, never an impersonation.

Compose two personalities

Combine a person with a nationality — one of each — for a layered personality:

With /goza linus-torvalds cubano — explaining a git merge:

Mira, sin rodeos: un merge coge dos ramas y las une en una. Git combina los cambios solo, y donde dos tocaron la misma línea te para en seco —eso es un conflicto—. No lo resuelvas a la loca: mira el código, entiende qué cambió cada lado y quédate con lo que de verdad funciona. Después, un commit limpio y palante.

Linus's direct, evidence-first judgment meets Cuban language and register — still the exact same technical answer underneath.

---

How it works

1. Install the personalities you want (one command, any supported agent).
2. Activate with /goza <name> — a nationality, a character, or one of each.
3. Ask anything. The personality stays on until you say modo normal.

Goza never changes the task, the tools, or the technical substance. It changes how the answer reads, not what the answer is.

---

Install

The fastest path uses the skills CLI — no clone, works across Claude Code, OpenCode, Codex, Cursor, and more:

npx skills add Ross-cripto/goza --skill goza --skill peruano --skill kirito -a claude-code -y

Preview everything available first:

npx skills add Ross-cripto/goza --list

Prefer cloning, a plugin with persistence hooks, or a manual copy? See INSTALL.md for every option.

---

Activate & compose

/goza cubano            # a nationality: language and local register
/goza kirito            # a character: temperament and framing
/goza kirito cubano     # compose one character + one nationality

• The nationality layer controls language and local register.
• The character layer controls temperament and framing.
• Combine one of each for a layered personality. Turn it all off with modo normal.

In agents without slash commands, just ask the agent to activate the cubano skill (or any name) from its installed SKILL.md.

---

The catalog

180+ personalities, each self-contained in its own SKILL.md:

• Nationalities — grounded in real language and register, added in reviewed regional batches so local expressions are authentic rather than mass-generated. Inventories: AMERICAS.md · WORLD.md
• Characters — people and figures with their own temperament and framing. Inventory: people/WORLD.md

Every profile respects regional, generational, and community variation, preserves your exact technical material, and always follows an explicit language request.

---

Design principles

• Substance is untouchable. Accuracy, tools, code, and reasoning never change.
• Grounded, not stereotyped. Personalities reflect real register and temperament — no caricature, no forced accent or slang.
• Portable core. The profiles are plain SKILL.md files and need no hooks. Adapters (a native /goza command, optional persistence hooks) are conveniences layered on top.
• Composable. One character, one nationality, or one of each — nothing more, so the personality stays coherent.

---

Add a personality

Add a directory under skills/people/ or skills/nationalities/ with a matching SKILL.md, README.md, and metadata.goza-type. Goza discovers it automatically — no changes to existing profiles. Start from docs/PROFILE_TEMPLATE.md and see CONTRIBUTING.md for the rules.

---

Project structure

skills/
├── nationalities/<name>/   SKILL.md + README.md   # language & local register
└── people/<name>/          SKILL.md + README.md   # temperament & framing
goza/                       the dispatcher (not a personality)
commands/ .claude/ .opencode/ hooks/               # optional per-agent adapters

---

Docs

• INSTALL.md — installation, adapters, and troubleshooting
• CONTRIBUTING.md — profile and adapter contribution rules
• docs/PROFILE_TEMPLATE.md — profile authoring template
• SECURITY.md — security boundaries and reporting
• RELEASING.md · CHANGELOG.md — releases and history

Validate the repository with npm test.

---

Distinct people. Helping you understand.