--- title: "AI Skill Design Practice Exercise for PMs" description: "Turn a recurring PM task into a reusable AI skill: design its context layer, write the skill, run it twice, and score consistency. A 90 minute exercise." canonical_url: "https://builderscamp.com/guides/challenges/building-your-ai-operating-system-reusable-ai-skill-design" date_published: "2026-09-16" date_modified: "2026-09-16" author: "Andre Albuquerque, Inês Lourenço" publisher: "Builders Camp" guide_class: "challenges" --- # AI skill design practice exercise for product managers **TL;DR:** This exercise has you design one working slice of a personal AI operating system: choose a recurring PM task, build the context layer an agent needs to do it reliably, write it as a reusable skill, run it twice on real input, and score whether the output is actually consistent enough to trust. ## The scenario Typing a fresh prompt every time you need a research summary or a PRD draft gets you a fresh, unpredictable result every time. An operating system approach is different: instead of typing prompts, you design a context layer that feeds an agent the right background information every time, write the task as a reusable skill, and let orchestration handle the repetition. The payoff is not a clever one-time answer, it is a dependable weekly output you no longer have to think hard about producing. The exercise asks you to build exactly one slice of that system, end to end, rather than a whole platform. You choose a real, recurring task you actually do, something structured enough to be worth systematizing, and take it from a vague habit to a documented, invokable skill with a defined context layer behind it, then check whether running it twice produces something consistent enough to actually rely on. ## What you are asked to do The exercise has four connected steps: - Choose one recurring product task you do repeatedly, one that is structured enough to be worth systematizing, such as research synthesis, PRD drafting, a weekly status update, or a competitive teardown. - Design the context layer the agent needs to do this reliably: product background, ideal customer profile, tone, constraints, and one concrete example of what good output looks like, structured as markdown sections you could drop into a context file. - Write the reusable skill itself, the instruction the agent follows every time, including the steps it takes, the inputs it expects, and the output format, generic enough to reuse the following week without rewriting it. - Run the skill on real input twice, score both runs on consistency, quality, and time saved, and name one specific change you would make to the context layer to make the output more reliable. ## What a strong answer covers The exercise's own objectives are the standard a strong submission has to meet: - Is the chosen task genuinely recurring and structured, rather than a one-off exercise dressed up as a repeatable workflow? - Does the context layer include a concrete example of good output, not just abstract background information the agent has no way to apply? - Is the written skill generic enough to invoke again on a different week's input without rewriting the instructions each time? - Do the two runs actually get compared against each other on consistency, not just evaluated individually for quality? - Does the named fix to the context layer address something specific that showed up across the two runs, rather than a general wish for better output? ## Skills this exercise practises Designing a context layer that an agent can reuse instead of relying on your memory of what background it needs each time. Writing a task as a generic, invokable skill rather than a prompt tailored to one specific moment. Evaluating whether an AI-assisted workflow is actually repeatable, not just impressive once. These map directly onto the bootcamp's own curriculum on context layer design, reusable PM skills, and automation with quality control. For the wider automation context this kind of skill often plugs into, see the [broken support automation exercise](https://builderscamp.com/guides/challenges/automate-workflows-with-ai-broken-support-automation) from Automate Workflows with AI, and for the underlying product-building groundwork, [how to build an AI product](https://builderscamp.com/guides/tools/how-to-build-an-ai-product). ## Which bootcamp this comes from This exercise is the practical challenge from [Building your AI Operating System](https://builderscamp.com/bootcamps/building-your-ai-operating-system), a two week bootcamp on Builders Camp with three live sessions covering context layer design, reusable PM skills, agent orchestration, and research synthesis workflows, taught by Inês Lourenço. Completing the practical challenge counts toward the bootcamp's completion requirement and its certificate, alongside the certification quiz. Builders Camp runs this bootcamp both live and self-paced, included with the Builders Camp Membership alongside every other bootcamp, track, and masterclass. If the recurring task you pick is a PRD, the [PRD template](https://builderscamp.com/guides/templates/prd-template) is a useful starting structure for the skill's expected output format. ## Frequently asked questions ### What does this AI operating system exercise ask you to build? One complete slice of a personal AI operating system: pick a recurring product task, design the context an agent needs to do it reliably, write it as a reusable skill, then run it twice and score the consistency, quality, and time saved. ### How is this different from writing a good one-off prompt? A one-off prompt gets a one-off result. This exercise is about designing something you can invoke again next week with the same structure and get a comparably good result, which needs a defined context layer and a generic enough skill definition, not just clever wording. ### What kind of recurring task works best for this exercise? Something repetitive and structured enough to systematize: research synthesis, a first draft of a PRD, a weekly status update, a competitive teardown. A one-time, highly unusual task will not show you whether the skill is actually repeatable. ### Why does the exercise require running the skill twice? Because a skill that works well once could just be a lucky run. Running it on real input twice and comparing consistency, quality, and time saved is what tells you whether the design actually holds up, and what specifically to fix in the context layer if it does not. ### How long does the exercise take? About 90 minutes, rated advanced difficulty, inside a two week bootcamp with three live sessions. ### Does completing it count toward a certificate? Yes. Finishing the practical challenge counts toward completing the Building your AI Operating System bootcamp on Builders Camp, alongside the certification quiz, and the bootcamp issues a certificate on completion. ## Sources - [Builders Camp: Building your AI Operating System bootcamp page](https://builderscamp.com/bootcamps/building-your-ai-operating-system) ## How this guide was made Researched from Builders Camp's bootcamp, track and masterclass material and the sources listed on this page, drafted with AI, and fact-checked against every source cited.