name: structured-arguments description: "Use this skill when working with labeled thesis arguments — creating, referencing, or weaving definitions (D), views (V), arguments (A), alternative views (AV), counter-arguments (CA), rebuttals, or barriers (B) into the thesis. Trigger when the user mentions argument IDs, cross-references, or wants to build/extend the argument scaffold."
Structured Arguments Skill
Modeled after Belcak et al. (2025) "Small Language Models are the Future of Agentic AI" — a position paper that labels every claim, argument, and counter-argument with short IDs and cross-references them throughout.
Element Taxonomy
Every element gets a short ID prefix + number and explicitly declares what it supports or attacks.
| Prefix | Type | Purpose | Example |
|---|---|---|---|
| D | Definition | Working definition anchoring a key term | D1: "Saturn is a zero-configuration..." |
| V | View | Core thesis claim (the position) | V1: "Zero-config AI provisioning is feasible" |
| A | Argument | Evidence or reasoning supporting a view | A1: "Six implementations prove feasibility" → V1 |
| AV | Alternative View | A counterposition from literature or practice | AV1: "Trust-the-network violates zero-trust" → disputes V3 |
| CA | Counter-Argument | Specific reasoning backing an alternative view | CA1: "BeyondCorp requires device-level auth" → supports AV1 |
| B | Barrier | Practical obstacle to adoption (not a logical counter) | B1: "No ecosystem adoption yet" |
| S | Step | Procedural/algorithmic step in a process | S1: "Discover services via mDNS" |
Rules
- Every A must link to at least one V. State it explicitly: "This argument supports view V1."
- Every AV must identify which V it disputes. "This alternative view challenges V2."
- Every CA must link to an AV. Counter-arguments exist in service of an alternative view.
- Rebuttals reference by ID. "We address counter-argument CA1 with arguments A3 and A7."
- Cross-references use IDs inline. "As shown in argument A2..." not "As shown previously..."
- IDs are stable. Once assigned, an ID never changes meaning. New elements get new numbers.
Registry
The argument registry lives at moons/claims/argument-registry.md. It is the single source of truth for all labeled elements. Before creating a new element, check the registry to avoid ID collisions.
Registry Format
## Definitions
- **D1**: [term] — [one-line definition]
## Views (Thesis Claims)
- **V1**: [one-line claim statement] — discussed in [chapters]
- **V2**: ...
## Arguments
- **A1**: [one-line summary] → supports **V1** — [evidence type: existence proof | literature | measurement | comparison]
- **A2**: ...
## Alternative Views
- **AV1**: [one-line counterposition] → disputes **V1** — [source: literature | practice | reviewer]
## Counter-Arguments
- **CA1**: [one-line reasoning] → supports **AV1** — [source]
## Barriers
- **B1**: [one-line obstacle] — [status: active | diminishing | addressed]
LaTeX Integration
Use these commands in thesis.tex for labeled arguments. Define them in the preamble:
% Structured argument labels
\newcommand{\argref}[1]{\textbf{#1}} % inline reference: \argref{V1}
\newcommand{\argdef}[2]{\noindent\textbf{#1}\quad #2} % definition block: \argdef{A1}{SLMs are...}
Rendering Patterns
Definition block (like the paper's WD1, WD2):
\begin{description}
\item[\argref{D1}] A \emph{Saturn service} is any AI endpoint advertised via \texttt{\_saturn.\_tcp.local.} mDNS records.
\end{description}
View statement (like V1, V2, V3):
We contend that:
\begin{description}
\item[\argref{V1}] Zero-configuration network protocols can provision AI services without end-user configuration;
\item[\argref{V2}] network-provisioned AI reduces total configuration effort compared to per-user manual setup;
\item[\argref{V3}] the security trade-offs of broadcast AI discovery are documented and addressable.
\end{description}
Argument block (like A1, A2):
\subsubsection{Six implementations prove feasibility}
\begin{description}
\item[\argref{A1}] Saturn's six independent client implementations demonstrate that mDNS/DNS-SD can provision AI services across languages and platforms. This argument supports view~\argref{V1}.
\end{description}
Cross-reference in prose:
As established in argument~\argref{A1}, the protocol works across five mDNS libraries.
We address counter-argument~\argref{CA1} in Section~\ref{sec:rebuttals}.
Workflow
Creating a New Element
- Read
moons/claims/argument-registry.mdto find the next available ID. - Draft the element with its one-line statement and linkage.
- Add it to the registry.
- If the element belongs in a specific chapter, note the chapter in the registry entry.
- When writing the chapter content, use the LaTeX patterns above.
Weaving Arguments Into Prose
When writing a thesis section that involves arguments:
- Identify which V/A/AV/CA elements are relevant to the section.
- Introduce the view (V) first, then present supporting arguments (A) with evidence.
- Present alternative views (AV) honestly, then their counter-arguments (CA).
- Rebut with cross-references to your own arguments.
- Every cross-reference uses the ID: "see argument A3" not "see above."
Mapping Saturn's Existing Claims
The three existing claims map directly:
| Existing | New ID | Statement |
|---|---|---|
| claim-1 | V1 | Zero-config AI provisioning is feasible via mDNS/DNS-SD |
| claim-2 | V2 | Network-provisioned AI reduces total configuration effort |
| claim-3 | V3 | Security trade-offs are known and addressable |
The evidence in moons/claims/claim-{1,2,3}.md and moons/claims/evidence-matrix.md becomes the source material for specific A arguments.