Can an AI Agent Really Follow Your Team's Testing Rules?

Ask an AI coding agent to write Playwright tests, and the first run might follow your team's conventions while the next one uses different selectors, ignores your fixtures and breaks your project's structure. You could paste the same instructions into every prompt. An agent skill writes them down once, in a form the agent can reuse. An agent skill is a reusable package of instructions that helps an AI agent do one kind of task. In the open Agent Skills format it is a folder with one required file, SKILL.md, which starts with a name and a description of when to use the skill, followed by the instructions. The folder can also hold scripts, reference documents and templates. Agents load skills in three stages. At discovery they read only each skill's name and description. At activation, a matching task pulls the full instructions into context. At execution, the agent follows them and opens bundled files only if it needs to. That is why a vague description matters: it decides whether the skill is used at all. The episode also compares a skill with a prompt and a tool, walks through a six-step skill for Playwright tests, sets out what to test in a skill, and explains why a skill is not training and why a skill from outside your team needs reviewing. ⏱️ Chapters 0:00 Same request, different tests 0:29 What the AI Tester Dictionary is 0:39 A reusable set of instructions 1:19 Inside a skill 1:55 Three stages 2:39 Prompt, skill and tool 3:18 A skill for Playwright tests 4:13 Testing a skill 4:53 A skill is not training 5:29 What to hold on to 🔑 Key takeaways • An agent skill is a reusable set of instructions and resources that helps an AI agent do one kind of task in a defined way • The agent reads each skill's name and description first and loads the full instructions only when a task calls for them, so the description decides when the skill is used • A skill describes the approach and a tool provides the capability: when a run goes wrong, work out which one failed • A repeatable process can still produce wrong results, so test the skill's instructions, its outputs and what the agent claims it did • A skill does not train the model or grant new permissions, and a skill from an untrusted source can carry harmful instructions, so review it like code 📚 In this series #001 — Artificial Intelligence #002 — Machine Learning #003 — Deep Learning #004 — Generative AI #005 — Large Language Models #006 — Transformer #007 — Diffusion Model #008 — Foundation Model #009 — Token #010 — Context Window #011 — Prompt #012 — System Prompt #013 — Prompt Engineering #014 — Temperature #015 — Top K and Top P #016 — Hallucination #017 — Embeddings #018 — Vector Database #019 — Semantic Search #020 — Retrieval-Augmented Generation #021 — Chunking #022 — Reranking #023 — Knowledge Graph #024 — Multimodal AI #025 — Function Calling #026 — Structured Output #027 — Grounding #028 — Fine-Tuning #029 — Context Engineering #030 — AI Agents #031 — Tool Calling #032 — Agentic AI #033 — Multi-Agent Systems #034 — Agent Orchestration #035 — Agent Memory #036 — Agent Skills ← you are here ▶️ Previously #035 — Agent Memory: what an agent keeps from one step or session to the next, and how to test it.    • Can You Trust What Your AI Agent Remembers...   ▶️ Next in the series #037 — MCP Servers: how AI applications connect to the tools and data that testers already use. 🔔 Subscribe for practical AI and software testing content. #SoftwareTesting #AIForTesting #AITesterDictionary #AgentSkills #AIAgents #Playwright #QAEngineer #SDET #AITesting #TestAutomation #AgenticAI #QualityEngineering #SoftwareQuality FOLLOW ME 📸 Web: https://softwaretestingtrends.com/ Discord:   / discord   LinkedIn:   / ambysan   Twitter:   / ambysan   Facebook:   / softwaretestingtrend   Medium:   / ambysan   BLOGGING 📝 Software Testing Trends: https://softwaretestingtrends.com/ TechJournal: https://ambreenkhan.dev/ ABOUT ME 👩‍💻 https://ambreenkhan.netlify.app/