Sergio Duran

OAuth 2.1 + Bearer HTTP front-end for gbrain serve (stdio MCP). PKCE + DCR + refresh tokens. Connected: ChatGPT, Claude.ai, Codex CLI, Claude Desktop, OpenClaw, Hermes. Tailscale Funnel compatible.

Connect a cloud Hermes gateway to a local desktop Hermes worker over Tailscale

🧠 One command to orchestrate your entire AI agent stack. /gbrain sync canonically updates Claude Code + OpenClaw + HERMES + GBrain. 25-section dashboard auto-verified by lie-detector. Layer 19 AWS Infra. Anti-overwrite SOUL.md guard. 60+h of production debugging distilled. Zero downtime.

ARCHIVED 2026-05-17 - moved to durang/skills (shared/gbrain). 19 layers + AWS Infra + sync orchestrator + lie-detector + compound engine.

Motor de motion design por codigo: numpy + PIL directo a ffmpeg. Sin After Effects, sin navegador headless, sin servicios de pago. Conectores 100% locales.

edit-video v3.13.0 · Edita y crea video con cualquier agente. Niveles: 1 Recorte · 2 Editorial · 3 Estudio (motion) · 4 Director. Desde cero: audio, guion o idea → .mp4 + .html. Clipper: mejores momentos con puntaje, lote, miniaturas, sigue al que habla. Guía de uso → GUIA.md

Video largo a clips verticales con subtitulos quemados. whisper + ffmpeg, sin suscripciones. La eleccion de momentos NO se automatiza: es criterio.

Skill /prompt para Claude Code: convierte peticiones en bruto en prompts blindados, con bloque de fusion

Claude/Codex PROMPT SKILL — /prompt: convierte ideas en bruto en prompts blindados (fases, verificación en vivo, reglas duras, entregable). No ejecuta: tú decides.

Setup-from-zero guide for the Meta Agentic Bridge stack — paste to Claude Code for guided install

Adversarial multi-round bug audit skill for Claude Code with self-learning pattern catalog. 5 rondas + triage obligatorio + patterns que compound knowledge across sessions.

Claude Code skill: connect any new machine to your Claude Code on EC2 via SSH + Tailscale + tmux. One curl-line install. 3 minutes setup.

Two AI agents on one WhatsApp number. OpenClaw captures silently → GBrain. Hermes responds with per-contact .md profiles, 3-phase memory (read→remember→save), GBrain hot memory (auto fact extraction v0.31+), live state, MCP tools, approval flows, scope. Maximum restriction default. 16-section visual dashboard. v5.8

Save-to-brain macro for ALL Claude clients (CLI, Desktop, web, mobile). Companion to garrytan/gbrain. Triggers on Spanish + English save phrases — agent infers entities, decisions, originals, and links from the conversation. Anti-hallucination contract included.

Connect any Claude (Code, Desktop, web, Cowork, mobile) to your GBrain in 60 seconds — one-command installer with HTTP wrapper, Tailscale Funnel, OAuth 2.1, and ambient capture.

Just One Brain. Your AI agent's memory, skills, and protocols — synced across every machine and every harness (Claude Code, Cursor, Windsurf, OpenClaw) via plain git. No SaaS, no lock-in.

🔍 OpenClaw Command Center — Infrastructure scanner & optimizer for Claude Code. Two skills that give you full visibility and actionable insights into your OpenClaw installation: system health, agents, plugins, memory, cron, security, disk usage — all in one dashboard. Built to run from any Claude Code session with zero config.