Skip to main content

Capability ยท Agentic engineering

Claude Code as an engineering platform, not a chat window.

In short

Claude Code becomes far more useful when a team packages its knowledge into skills, plugins and MCP servers that every session inherits. I build those: versioned Claude Code plugins with tests, agent skills for repeatable work, sub-agents for review, and MCP servers that connect Claude to business systems such as Xero and Microsoft Ads, with permissions kept tight.

I do this work as AI & Automation Lead at Dexterous Group in Melbourne. This page shows the capability; it is not an offer of personal services.

The method, step by step.

  1. Capture the method, not the prompt

    A skill holds how a task is done, with its checks. Plugins bundle skills, commands and hooks so a whole team works the same way.

  2. Version and test it

    Plugins get a version number, a changelog, tests and a release script, like any other software.

  3. Connect systems through MCP

    MCP servers expose only the actions a task needs, read-only by default, with the credentials kept outside the model.

  4. Keep a human in the loop

    Anything that writes to a business system asks first, and every action is logged.

Where I have done it.

Anonymised: no client, employer product or confidential detail.

Problem
A finance team repeated the same checks and coding by hand every week.
What I did
Built a private Claude plugin marketplace of 11 versioned finance plugins with tests and a release script.
Result
Repeatable finance work packaged so every session follows the same method.
Stack
Claude Code plugins and skills, Python
Problem
Moving websites off WordPress was slow and error-prone.
What I did
Built a CLI and 9 Claude skills that assess, capture, transform, verify and deploy a site as a static, agent-readable site.
Result
Several live sites moved, with every old URL checked before cutover.
Stack
Node.js, Claude skills, Cloudflare Pages and Workers, GitHub Actions
Problem
Accounting data and ad data were locked behind separate dashboards.
What I did
Built and documented MCP servers for Xero and Microsoft Ads.
Result
Claude can read and analyse that data in plain language, within set permissions.
Stack
MCP, TypeScript, Xero API, Microsoft Advertising API

What I work with.

  • Claude Code
  • Claude plugins and marketplaces
  • Agent skills
  • Sub-agents and hooks
  • MCP servers
  • TypeScript and Python

Frequently asked.

What is a Claude Code plugin?

A plugin is a package of commands, agent skills, hooks and MCP servers that extends Claude Code. Teams install plugins from a marketplace so everyone works with the same tools and methods.

What is the difference between Claude skills and plugins?

A skill is one packaged capability: instructions, scripts and checks for a task. A plugin can bundle several skills with commands, hooks and MCP servers, and is versioned and installed as a unit.

What is an MCP server?

MCP (Model Context Protocol) is an open standard for connecting AI models to tools and data. An MCP server exposes a system, such as an accounting platform, as actions an AI model can call within permissions you set.

Is Claude Code safe for business use?

It can be, with controls: limit which commands and files it may touch, keep secrets out of its reach, require approval for writes, and review its changes like any other code.

Also on this site.

For project work, contact Dexterous Group. For speaking, media or a peer conversation, email mail@amjid.au or message me on LinkedIn.