Skills

Custom AI agent instructions — a collection of thinking partners and methodological guides that shape how the agent reasons, writes, and teaches across domains.

Wait, how the workflow should be?

Say doing a project, ask to the LLM with some architecture, and it should be the one that call the partners, with a specific context window, with a specific question, or directly with a note.

So the one which i talk to is always the orchestrator, generally talking.

Of course that there is going to be session that I know that I want to advance with a project, but I would love that while doing a project, somehow the orchestrator call to other partners

Structure

skills/
├── PARTNERS/                    # Domain-specific thinking partners (agent personas)
│   ├── bloch.md                 # Computational Physics & DFT
│   ├── cauchy.md                # Real Analysis
│   ├── cayley.md                # Linear Algebra
│   ├── euler.md                 # Mathematical Methods for Physics I
│   ├── feynman.md               # Quantum Computing
│   ├── fourier.md               # Mathematical Methods for Physics II
│   ├── kirchhoff.md             # Analog Electronic Circuits
│   ├── lagrange.md              # Classical Mechanics
│   ├── linus.md                 # Linux & Computer Architecture
│   ├── optics.md                # Optics (Fresnel)
│   ├── planck.md                # Modern Physics
│   └── shannon.md               # Deep Learning & LLMs
│   └── connections/             # Cross-partner connection notes
│       ├── bloch-connections.md
│       ├── cayley-connections.md
│       ├── feynman-connections.md
│       ├── fourier-connections.md
│       ├── kirchhoff-connections.md
│       ├── lagrange-connections.md
│       └── planck-connections.md
└── SOULS/                       # Methodological guides
    ├── zettelkasten.md           # Zettelkasten note creation (general)
    └── zettelkasten_saas.md      # Zettelkasten adapted for SaaS due diligence

PARTNERS — Thinking Partners

Each partner is a curated AI persona with a distinct voice, philosophical stance, and domain expertise. They define not just what the agent knows but how it thinks — what questions it asks, what approximations it challenges, and how it connects ideas across Jorge's interdisciplinary landscape (physics, math, ML, systems).

Partner	Domain	Named After	Core Question
Bloch	Computational Physics & DFT	Felix Bloch	What approximation is being made, and what breaks when it fails?
Cauchy	Real Analysis	Augustin-Louis Cauchy	What is the intuition before the definition, and what is the key move in the proof?
Cayley	Linear Algebra	Arthur Cayley	What vector space are we in, and what's the basis?
Euler	Methods I	Leonhard Euler	What physical system is this equation modeling, and why is this the right tool?
Feynman	Quantum Computing	Richard Feynman	What physical resource — superposition, entanglement, or interference — is being exploited?
Fourier	Methods II	Joseph Fourier	What basis are we in, and why is it the right one?
Fresnel	Optics	Augustin-Jean Fresnel	Why does this model work, and where does it quietly fail?
Kirchhoff	Analog Circuits	Gustav Kirchhoff	What does the scope say? Trust the measurement over the calculation.
Lagrange	Classical Mechanics	Joseph-Louis Lagrange	What are the generalized coordinates, and what do the symmetries conserve?
Linus	Linux & Computer Arch	Linus Torvalds	How does this actually work under the hood, and what leverage does that give you?
Planck	Modern Physics	Max Planck	What classical assumption broke here, and what replaced it?
Shannon	Deep Learning & LLMs	Claude Shannon	What is the mathematical object here, and what physics maps onto it?

Cross-Partner Connections

Each partner is initially agnostic of the others. When a connection to another domain is needed, Jorge references a connection note from PARTNERS/connections/. These notes describe how one partner's domain relates to another's — allowing the current partner to connect their knowledge without having hardcoded knowledge of the other agents.

Connection	Link
Bloch ↔ Shannon	bloch-connections.md
Fourier ↔ Bloch, Planck, Euler	fourier-connections.md
Lagrange ↔ Planck, Euler/Fourier, Bloch	lagrange-connections.md
Planck ↔ Fourier, Bloch, Lagrange	planck-connections.md
Kirchhoff ↔ Planck/Bloch, Fourier, Lagrange	kirchhoff-connections.md
Feynman ↔ Planck, Shannon, Fourier, Bloch, Linus	feynman-connections.md
Cayley ↔ Planck, Shannon, Fourier, Lagrange, Feynman, Kirchhoff	cayley-connections.md

Shared Design Philosophy

Every partner follows the same template:

• Personality — a voice and intellectual stance, not a neutral tutor
• Scope — bounded domain expertise with explicit edges
• How to work with USER — concrete interaction patterns, question styles, and domain-specific heuristics
• Zettelkasten approach — how to write atomic, linked notes that capture the mechanism, not just the name

SOULS — Methodological Guides

Guidelines for how notes should be structured, written, and interconnected.

Guide	Purpose
zettelkasten.md	General Zettelkasten methodology — atomicity, self-containment, connectivity, emergent organization. Defines note types (literature, permanent, project), Obsidian conventions, and quality checks.
zettelkasten_saas.md	Adapted Zettelkasten for startup due diligence — research notes, thesis notes, business-model conventions ([!metric], [!risk], [!assumption]), and VC-pitch-oriented link thinking.

Usage

These files are meant to be loaded as context when the agent needs to adopt a specific thinking partner or follow a particular methodology. They set the voice, constraints, and quality bar for the agent's output in that domain.