Pietro Montaldo on Claude Plugins vs Skills

Pietro Montaldo recently shared something that caught my attention: "Claude Plugins > Claude Skills." He followed it with a promise that matters to a lot of operators who do not want another vague, overhyped AI post: "Here is my full tutorial for non-technical operators" and "This is not another AI slop guide."

That framing resonates because most people are not stuck on whether AI is useful. They are stuck on making it repeatable. Pietro also offered a clean mental model that explains why many teams feel like they are spinning their wheels with chat tools:

"Simple Chat -> Skills -> Plugins"

In this post, I want to expand on Pietro's ladder and make it actionable: what each rung really means, when to use it, and how to move from one-off prompting to packaged workflows that your team can actually run.

The real problem: repeating yourself is expensive

If you have ever said, "AI is helpful, but it takes time," you are not imagining it. The hidden cost is not the tool. It is the constant re-explaining:

• Repeating context (brand, audience, product details)
• Repeating standards (tone, formatting, do and do not lists)
• Repeating process (steps, QA checks, approval rules)
• Repeating handoffs (what goes into the doc, CRM, or analytics sheet)

When AI lives only as a chat box, every request starts from scratch. The outputs can be good, but the system is fragile. Pietro's point is that you should graduate from "chat" to "assets".

Simple Chat: powerful, but you pay the context tax

"Simple Chat = You explain everything every time." That is the tradeoff.

Simple chat is best when:

• The task is novel or exploratory (brainstorming, research outlines)
• The stakes are low (drafting internal notes)
• The workflow changes often (early stage experiments)

But simple chat breaks down when you need consistency. For example, a marketing operator might ask for a LinkedIn post, then ask again tomorrow, and get a different voice, structure, or set of claims. Nothing is saved except what is in your head.

Quick win: make your first reusable "brief"

Before you even touch skills or plugins, create a reusable brief you can paste:

• Who we are (one paragraph)
• Who we serve (personas)
• What we sell (offer + differentiators)
• Voice and tone (3-5 bullets)
• Proof sources allowed (case studies, pages, docs)
• Red lines (no promises, no sensitive claims)

This alone reduces prompt length and improves output quality. But it still depends on you pasting it.

Skills: the leap from prompts to repeatable behavior

Pietro summarized it simply: "Skills = You explain it once. Claude remembers." The deeper idea is standardization.

A skill (in plain language) is a saved, reusable operating procedure for the AI. You invest once, then run it many times.

Skills are best when:

• The task repeats weekly or daily
• You want the same structure every time
• Multiple people should get similar outputs

What a good skill contains

Think like an operator writing a checklist. A strong skill usually has:

1. Purpose: what the skill is for
2. Inputs: what the user must provide (topic, audience, links)
3. Process: the steps the AI must follow
4. Output format: template, sections, length, tone
5. QA rules: accuracy checks, citations, constraints

Key insight: A skill is a process, not a prompt.

Example skill: "Content repurposing with guardrails"

If you turn webinars into posts, your skill might:

• Ask for the transcript link or raw text
• Extract 5 key points
• Draft 3 post angles for different segments
• Produce 1 final post in your brand voice
• Include a compliance checklist at the end

Now you have consistent outputs without rewriting instructions every time.

Plugins: packaging multiple skills plus tools into one workflow

Pietro's final step is where most non-technical teams start to see compounding returns: "Plugins = Multiple skills + tools + workflows in one package."

The important addition is "tools." A plugin is not just a better prompt. It can connect the AI to actions or data sources so the workflow is not trapped inside a chat.

Plugins are best when:

• You need repeatable, multi-step workflows end to end
• The workflow touches external systems (docs, sheets, CRM, analytics)
• You want fewer manual copy-paste steps

What changes when tools enter the picture

Without tools, AI drafts and you execute. With tools, AI can help execute (within permissions).

Examples of plugin-style workflows a growth or marketing operator might want:

• Content workflow: generate outline -> draft -> create variations -> save to a content calendar
• Research workflow: pull competitor pages -> summarize positioning -> output a comparison table
• Lead workflow: enrich a list -> segment -> draft outreach messages -> log to a sheet

Even when the plugin does not directly push buttons in external apps, it can still bundle the entire operating system of the task: the instructions, the templates, the QA rules, and the handoff artifacts.

A practical path: move up the ladder without boiling the ocean

Pietro's "Simple Chat -> Skills -> Plugins" is a maturity model. Here is how I would apply it in a team setting.

Step 1: Identify one repeating task with clear ROI

Pick something that happens often and hurts when it is inconsistent:

• Weekly performance summary
• Customer interview analysis
• Campaign brief creation
• LinkedIn post drafting and editing

If it is not repeated, do not standardize it yet.

Step 2: Turn your best prompt into a skill

Use your top-performing chat exchange as raw material. Then rewrite it as a skill:

• Remove one-off context
• Add required inputs
• Add structure and QA rules
• Add examples of good output

Run it five times. Tighten it. Let someone else run it. Tighten it again.

Step 3: Bundle adjacent tasks into a plugin workflow

Once the skill is stable, look left and right:

• What happens before this step?
• What happens after this step?
• Where do people copy and paste?

That is your plugin opportunity: package the full workflow so execution becomes simpler than improvisation.

If using the workflow feels easier than doing it manually, adoption takes care of itself.

Common mistakes when teams jump to plugins too early

Pietro's post is aimed at non-technical operators, which is exactly the group that gets burned by overbuilding. A few pitfalls to avoid:

1. Automating chaos: If your process is unclear, the plugin will scale confusion.
2. No inputs spec: People cannot use a workflow if they do not know what to provide.
3. No QA layer: Without checks, you get fast output and slow cleanup.
4. Too broad: One plugin that does everything usually does nothing well.

Start narrow, get trust, then expand.

Why this matters: consistency is the real productivity unlock

The biggest takeaway from Pietro Montaldo's framing is not that plugins are shiny. It is that they reduce variability.

• Simple chat gives you output.
• Skills give you consistency.
• Plugins give you systems.

In operations, systems win. They let you train new teammates faster, produce higher-quality work under time pressure, and reduce the cognitive load of "figuring it out again." That is what makes AI feel like leverage instead of another tab.

---

This blog post expands on a viral LinkedIn post by Pietro Montaldo, I build and share AI tools and playbooks actually useful for non-techies | DM for custom trainings or join my next AI-agent building course on Maven. View the original LinkedIn post →