"How I turned practical workflows, technical knowledge and creative problem-solving methods into a portable collection of AI agent skills."

Introduction

We've become accustomed to interacting with AI through prompts.

We describe what we want, provide context, receive an answer, and repeat the process whenever we need something similar.

But what happens when we stop treating every interaction as a separate instruction?

What if we could package a particular way of approaching a problem into a reusable capability?

This is where agent skills become interesting.

Rather than repeatedly explaining how we want an AI agent to approach a task, we can give it structured instructions, methodologies and supporting resources that help guide its behaviour.

I started exploring this idea through a project called BOONS.

1. The Problem With Repeating Yourself

A capable AI model doesn't automatically mean a consistent workflow.

The same model can approach similar tasks differently depending on the context, instructions and information supplied.

This creates an interesting engineering problem:

  • How do we preserve useful workflows?
  • How do we make instructions reusable?
  • How do we transfer capabilities between different agents?
  • How do we encourage more consistent approaches to specialised tasks?

The answer isn't necessarily another enormous system prompt.

Sometimes, what we need is a smaller, focused capability that can be called upon when the task requires it.

2. What Are Agent Skills?

Agent skills are structured packages of instructions and resources that teach compatible AI agents how to approach particular tasks.

A skill can define:

  • Its purpose and intended use.
  • The workflow an agent should follow.
  • Boundaries and constraints.
  • Relevant supporting documentation.
  • Verification procedures.

The important distinction is that a skill is not a new AI model.

It is a way of organising and delivering specialised instructions and knowledge to an existing agent.

This makes skills particularly interesting for developers who want to build reusable workflows without having to train a model from scratch.

3. Introducing BOONS

BOONS is an open-source collection of portable AI agent skills developed under TagzAuthor.

Rather than concentrating on one discipline, I wanted to explore how different areas of knowledge could be packaged into individual capabilities.

The collection currently contains six skills:

Cybersecurity

Security Methods

Evidence-led defensive security analysis, with an emphasis on authorization, scope, uncertainty and remediation.

Bug Bounty Recon

Scope-first bug bounty research and finding triage.

Coding

Craft and Vibe

Creative software engineering with deliberate verification.

Creative Systems

Systems Synthesis

Developing original systems from higher-level patterns and ideas.

Voice and Edit

Refining written material while preserving facts and the author's intended meaning.

Technical Analysis

REA Amplify

Authorized artifact analysis that grounds and supports another skill.

Each skill has a focused responsibility rather than trying to make one giant instruction set responsible for everything.

4. Why Specialisation Matters

Consider a security review.

A general-purpose AI agent might understand cybersecurity concepts, but that doesn't automatically mean it will consistently distinguish between:

  • Verified observations.
  • Inferences.
  • Assumptions.
  • Unconfirmed vulnerabilities.
  • Actions that require explicit authorization.

A specialised security skill can establish a more deliberate process for handling these distinctions.

The same principle applies to software development, writing and systems design.

Specialisation gives us a way to organise how an agent approaches a problem, instead of relying entirely on whatever instructions happen to be in the current conversation.

5. Portability: Building Beyond One AI Platform

One of the interesting aspects of the Agent Skills format is its portability.

Instead of tying every workflow to one particular AI product, compatible skills can be used across different agent environments.

This creates opportunities for developers to build collections of reusable capabilities.

A skill can be developed, tested, improved and shared without requiring the creator to build an entirely new AI application around it.

That changes the economics of experimentation.

You can focus on the capability itself rather than rebuilding the surrounding infrastructure every time.

6. From an Open Collection to Specialised Products

BOONS is currently distributed as one complete free package.

That is intentional.

I wanted the collection to be accessible, usable and open to experimentation.

But there is another direction worth exploring.

Individual capabilities can evolve into more specialised products with deeper methodologies, supporting documentation, structured workflows and more demanding validation.

For example, a basic defensive security methodology can develop into a much more comprehensive security assurance system.

The distinction should not simply be the number of instructions.

It should be the additional capability, structure and reliability that the specialised product provides.

7. The Bigger Picture

I find the interesting opportunity in agent skills to be the separation of intelligence from the way we package and apply it.

The model provides the underlying reasoning capability.

The skill provides specialised instructions and a repeatable approach.

The surrounding tools provide access to the environment in which the work happens.

Together, these components create opportunities to build more specialised AI workflows.

We are moving beyond asking what a model can answer.

We can also ask:

What capabilities can we build around it, and how reliably can those capabilities be reused?

Conclusion

BOONS started as an exploration of reusable AI agent capabilities.

It has become a collection of six skills spanning cybersecurity, coding, creative systems and technical analysis.

For me, this is one of the more interesting directions in AI development: building small, focused capabilities that can be combined, improved and shared.

The future of AI agents won't be determined by models alone.

It will also depend on the systems, tools, skills and methodologies we build around them.

More of My Work

The whole library: TagzAuthor

Explore BOONS

BOONS is a collection of AI agent skills you can bring into your AI workspace. Choose whichever access route works best for you.

GitHub — Open-source access

FlowArchitect895/Boons

Explore the collection, download the skills and add them to your own AI workspace.

Agensi — Cloud-based access

Explore BOONS on Agensi

Access the skills through Agensi's marketplace and bring them into your supported AI workspace.

Explore them, put them to use, and let me know how your experience goes. I'd love to hear your feedback: what worked, what didn't, and what you ended up creating with them.

Support the work: ko-fi.com/tagzauthor