Herding AI cats: What rules do you include in your Instructions to control your ai-assisted development?
26 Oct 2026
AI can churn out code, dead, dead fast, but that doesn't mean it's good or fit for purpose.
I've had success building applications with AI-assisted development, and I'm always keen to improve my workflow. The tester in me wants to know things. I want to know what's happening under the hood; I want feedback; I want to know how we can add some predictability to our AI output. Naturally, I'm looking at guardrails and using evals, etc., but I want to know what you're experimenting with.
I've found the structure of https://github.com/addyosmani/agent-skills/ helpful for learning how to keep instruction files small and call on things when relevant.
Most platforms allow you to give specific instructions and set boundaries through AGENT.md or CLAUDE. md files. Skills usually include boundaries and verification steps to give you some confidence.
Since you can add instructions at multiple levels, what instructions are you adding, and where? How do you check that your instructions are being followed, and what do you do when they aren't?
New to this, curious, or wondering what I'm on about? Add your questions too!
I've had success building applications with AI-assisted development, and I'm always keen to improve my workflow. The tester in me wants to know things. I want to know what's happening under the hood; I want feedback; I want to know how we can add some predictability to our AI output. Naturally, I'm looking at guardrails and using evals, etc., but I want to know what you're experimenting with.
I've found the structure of https://github.com/addyosmani/agent-skills/ helpful for learning how to keep instruction files small and call on things when relevant.
Most platforms allow you to give specific instructions and set boundaries through AGENT.md or CLAUDE. md files. Skills usually include boundaries and verification steps to give you some confidence.
Since you can add instructions at multiple levels, what instructions are you adding, and where? How do you check that your instructions are being followed, and what do you do when they aren't?
New to this, curious, or wondering what I'm on about? Add your questions too!
Christopher Chant
Business Growth & Agility Guide
He/Him
Christopher Chant is a determined and passionate test professional with experience across multiple domains. He has learned to embrace all parts of the development lifecycle as learning opportunities: working in business analysis, development, testing and coaching roles in an attempt to help teams grow and deliver. These days Christopher is looking at ways that the Satir Model and Cynefin can be used to aid teams and organisations make safe changes.
When not testing, Christopher spends his time running (not often enough), bouldering, travelling all over the country to watch Nottingham Forest F.C. lose (occasionally they win), jealously looking at other people's dogs and playing board games.
Open To
Teach
Speak
Mentor
Podcasting
Write
CV Reviews
Sign in
to comment
There are countless ways to show up and contribute. Which one will you try today?
Explore MoT
Complex workflows, flaky tests, fragmented tools, and AI-driven pressure are making it harder for QA teams to keep up.
Turn AI curiosity into engineering confidence.
Debrief the week in Quality via a community radio show hosted by Simon Tomes and members of the community
