Roberto Manuel Jara Peche

I build open-source AI agent systems that remember context, create useful outputs, and verify what matters.

GitHub · LinkedIn · YouTube · Instagram

---

What I build

My work sits at the intersection of agent orchestration, portable skills, memory, creative tooling, and evidence-based verification. The goal is practical: make AI systems more useful in real workflows without hiding uncertainty or breaking the user's control.

Pillar	Project	What it contributes
Remember	Wife	Local-first, persistent, auditable memory for agent workflows.
Create	ShotPilot	A portable image-generation skill that turns intent into structured, production-ready prompts and outputs.
Verify	Wonder Woman	Adversarial multi-agent verification for claims, decisions, and high-stakes work.
Reflect	Clear Mirror	Direct, evidence-grounded strategic reflection that exposes assumptions and leaves a prioritized plan.

Open-source focus

• Agent Skills that can move between harnesses instead of being trapped in one chat interface.
• Supervisor and verification patterns for work that needs evidence, review, and explicit uncertainty.
• Local-first tooling that keeps context inspectable and under the user's control.
• Practical AI interfaces for image generation, so