npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@chordandev/jorge-el-curioso

v0.2.0

Published

Install Jorge El Curioso advisory code review agents and skills for OpenCode, Gentle-PI, and Claude Code.

Readme

Jorge El Curioso — Multi-Agent Code Review (OpenCode, Gentle-PI & Claude Code)

Jorge El Curioso is an advisory, read-only multi-agent code-review workflow with deterministic evidence (Evidence v2). It provides multi-runtime support for OpenCode, Gentle-PI, and Claude Code.

  • En OpenCode: Instala 6 agentes de producción (jorge-el-curioso como primary coordinator + 5 subagentes).
  • En Gentle-PI: El orquestador es gentle-orchestrator (el Gentleman). Al activarse la skill jorge-el-curioso, el orquestador adopta el protocolo de revisión de Jorge y coordina 5 subagentes especializados (jorge-impact-mapper, los 3 especialistas de análisis y jorge-refuter) usando subagent_run.
  • En Claude Code: El agente principal (Claude) adopta la skill jorge-el-curioso como coordinador asesor y delega en 5 subagentes especializados (jorge-impact-mapper, los 3 especialistas con fan-out paralelo, y jorge-refuter) usando la herramienta nativa Agent.

Quick path

Running install in an interactive terminal without a target flag asks whether to install for OpenCode, Gentle-PI, Claude Code, or all environments. Explicit target flags and --no-interactive remain deterministic and prompt-free; the latter defaults to OpenCode. Installation never asks for individual model choices. After it finishes, the installer prints how to customize models and thinking levels: run jorge-el-curioso from a global installation (npm install -g @chordandev/jorge-el-curioso), or npx -y @chordandev/jorge-el-curioso when Jorge was used through npx.

Para Claude Code

npx -y @chordandev/jorge-el-curioso install --claude

Instala la skill en ~/.claude/skills/jorge-el-curioso/ y los 5 subagentes en ~/.claude/agents/.

Para Gentle-PI

npx -y @chordandev/jorge-el-curioso install --pi

Instala la skill en ~/.pi/agent/skills/jorge-el-curioso/ y los 5 subagentes en ~/.pi/agent/agents/.

Para OpenCode

npx -y @chordandev/jorge-el-curioso install --opencode

Instala los 6 agentes en ~/.config/opencode/agents/ y la skill.

Para todos los entornos

npx -y @chordandev/jorge-el-curioso install --all

Cambiar modelos de forma interactiva (Model Manager)

Podés cambiar o inspeccionar los modelos de Jorge en cualquier momento con una interfaz interactiva de terminal:

npx -y @chordandev/jorge-el-curioso
# O directamente si lo instalaste de forma global:
jorge-el-curioso

Al iniciarse, el asistente te preguntará qué entorno querés configurar (OpenCode, Gentle-PI o Claude Code). Si elegís un entorno que aún no está instalado en tu máquina, te informará que no está instalado.

  • Ver y editar agentes individualmente: con sugerencias rápidas de la familia OpenAI (gpt-6-astra, gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna).
  • Aplicar Presets recomendados (Variantes OpenAI):
    • 1. OpenAI Frontier: GPT-6 Astra (Refuter, Seguridad) + GPT-5.6 Sol (Coordinador, Resiliencia) + GPT-5.6 Terra (Correctness) + GPT-5.6 Luna (Mapper).
    • 2. OpenAI Full Astra: GPT-6 Astra en Coordinador, Especialistas y Refuter (máxima deducción para PRs hipercríticos).
    • 3. OpenAI 5.6 Sol (Flagship): GPT-5.6 Sol en Coordinador, Especialistas y Refuter; Luna en Mapper.
    • 4. OpenAI 5.6 Terra (Equilibrado): GPT-5.6 Sol en Coordinador y Refuter; Terra en Especialistas; Luna en Mapper.
    • 5. OpenAI 5.6 Luna (Rápido y Ágil): GPT-5.6 Luna en Coordinador y Especialistas; Terra en Refuter; Luna en Mapper.
    • 6. OpenAI Base Luna: GPT-5.6 Luna en todos los agentes.
  • Asignar un mismo modelo a todos de una sola vez.
  • Guardado automático directo en ~/.config/opencode/opencode.json y jorge-el-curioso.json.

También podés pasar flags no interactivos:

npx -y @chordandev/jorge-el-curioso --model jorge-refuter=openai/gpt-6-astra --no-interactive

Aislamiento de revisiones con Git Worktrees (Automatizado)

Cuando querés revisar un PR o commit que no es la rama en la que estás trabajando (o si tenés cambios locales sin commitear), Jorge exige que el código esté limpio y congelado en ese commit para evitar falsos positivos producidos por mezclar archivos de tu checkout actual con los del PR.

Podés preparar y validar ese entorno aislado en un solo comando:

# Creá un worktree aislado para el PR #100:
jorge-el-curioso worktree 100

# O pasando la URL completa del PR o un commit SHA:
jorge-el-curioso worktree https://github.com/Team-Origen/origen/pull/100
jorge-el-curioso worktree 0a2ca232070bc478fe2cb8105fd392aef109d429 ../mi-worktree-review

El comando automáticamente:

  1. Resuelve el commit inmutable del PR (vía gh o git).
  2. Descarga el commit si falta en tu repo local.
  3. Crea un worktree secundario en ../<repo>-pr-<id>-review en estado detached.
  4. Valida los 4 invariantes de Jorge: commit exacto, estado detached, árbol limpio y repositorio válido.
  5. Te indica cómo entrar y cómo eliminarlo cuando termines.

Comandos de gestión de worktrees

# Listar worktrees activos:
jorge-el-curioso worktree list

# Eliminar un worktree cuando terminás la revisión:
jorge-el-curioso worktree remove ../origen-pr-100-review

Herramientas deterministas incorporadas (Evidence v2)

Inspirado en los aprendizajes de producción de CodeRabbit, Jorge el Curioso ahora prioriza herramientas deterministas de primer orden sobre lecturas puramente estadísticas o conjeturas de LLM:

  1. LSP Diagnostics & Navigation: Consulta activamente compiladores y Language Servers (elixir-ls, TypeScript, etc.) para capturar errores de tipo, advertencias de sintaxis y call hierarchies.
  2. ast-grep (ast-grep / sg): Búsqueda estructural y sintáctica en el AST para patrones de código, call sites y anti-patrones de seguridad.
  3. Minería de Co-commits en Git: El Impact Mapper analiza automáticamente el historial reciente (git log --oneline --name-only) para detectar archivos históricamente acoplados que hayan sido omitidos en el PR.
  4. Logical Cohorts (Change Stack): Ordena la lectura en capas de dependencia lógica (Esquemas/Contratos → Dominio/Core → API/Límites → Tests) en lugar de una lista alfabética de archivos.
  5. Linters y Scanners de Sólo Lectura: Permisos habilitados para ejecutar linters del proyecto (mix compile, mix credo, mix sobelow, shellcheck) y extraer violaciones con receipts comprobables.
  6. Matriz Effort vs. Reward: El Refuter y los candidatos clasifican cada problema confirmado según su esfuerzo y beneficio (Quick wins vs. Heavy lifts).

The installer only deploys Jorge assets and, when explicitly requested, Jorge model settings. It never detects, installs, or configures language servers, linters, formatters, scanners, package managers, or Homebrew.

Optional project tooling

Jorge does not install or configure development tooling. Choose it for each reviewed project:

  • TypeScript: typescript, typescript-language-server, tsc --noEmit, and ESLint.
  • Elixir/Phoenix: ElixirLS, mix compile --warnings-as-errors, Credo, and Sobelow.

The installer repeats this checklist after a successful installation.

Model choices

The interactive flow asks for one model per production agent:

| Agent | Default model | | --- | --- | | jorge-el-curioso | openai/gpt-5.6-luna | | jorge-impact-mapper | openai/gpt-5.6-luna | | jorge-correctness-data | openai/gpt-5.6-luna | | jorge-transactions-resilience | openai/gpt-5.6-luna | | jorge-security-performance | openai/gpt-5.6-luna | | jorge-refuter | openai/gpt-5.6-luna |

For automation, repeat --model agent=model. This disables prompts:

npx -y @chordandev/jorge-el-curioso install \
   --model jorge-impact-mapper=openai/gpt-5.6-luna \
   --model jorge-refuter=openai/gpt-6-astra \
  --save-models

--save-models retains the resolved choices in ~/.config/opencode/jorge-el-curioso.json. It is ordinary JSON, intended to be edited by the user. On later runs, saved choices become the defaults before command-line overrides.

What the installer changes

| Location | Behavior | | --- | --- | | ~/.config/opencode/skills/jorge-el-curioso/ | Installs and safely upgrades the production skill and its two runtime references. | | ~/.config/opencode/agents/jorge-*.md | Installs and safely upgrades only the six production Jorge agent definitions. | | ~/.config/opencode/jorge-el-curioso.assets.json | Records the package version and SHA-256 of each last known package-managed asset. Do not edit this managed ownership manifest. | | ~/.pi/agent/skills/jorge-el-curioso/ | Installs the Gentle-PI coordinator skill and its references. | | ~/.pi/agent/agents/jorge-*.md | Installs only Jorge's five Gentle-PI subagent definitions. |

Upgrading installations created before the ownership manifest

Older releases did not record asset ownership and left stale Jorge prompts in place. On the first rerun with a manifest-aware release, each differing file at one of the nine exact Jorge-managed paths is copied next to itself as:

<asset>.jorge-el-curioso.<timestamp>.backup

The installer reports every exact backup path, deploys the current package template, and then writes the ownership manifest. Inspect or restore those backups if the old Jorge files contained custom changes. Later reruns preserve detected customizations instead of overwriting them. Restart OpenCode after every asset upgrade so the running process loads the new agent and skill prompts.

Safe checks

Preview without writing files:

npx -y @chordandev/jorge-el-curioso install --dry-run --no-interactive

The preview reads the current ownership state and reports planned installs, upgrades, legacy backups, preserved customizations, and manifest changes. It does not create directories, backups, temporary files, manifests, configuration, or software.

For package development:

npm test
npm run dry-run