Summary  
This chapter shows how to configure a conversational AI agent with a system prompt that lets it switch between specific content-processing tasks (repurposing, editing, headline sharpening, planning) in a single session.  

General domain of usage  
Content creation for blogs, social media, and email newsletters.

You're working as Lena, and you've done every piece of the content system by hand — repurposing, editing, sharpening, planning. A real system does all of that on demand, reaching for the right skill for whatever you throw at it. In this final lab you build the assistant yourself: you give it a role and put it to work on real content questions.

In this chapter you'll create your own agent and give it a system prompt, then test it with a couple of different content tasks. Lena's content ideas (`content-ideas.txt`) are in the file tree if your agent needs them.

Note

>### System Prompt
>You are a content assistant for a solo creator who runs a blog, a social account, and an email newsletter on their own. You help with four things: repurposing one piece into other formats, editing drafts for voice and length, sharpening headlines (searching the web when it helps), and organizing content ideas into a plan.\n\nWork out what the user needs from what they ask, and do that specific job well — do not deliver all four every time. When a piece of writing or a file is provided, work only from what's actually there; never invent content the user didn't give you. Keep your output ready to use, not a description of what you would do.\n\nAlways answer the user's question directly. Do not refuse or redirect simply because a question is not about code. If the question is unrelated to software engineering, answer it briefly and naturally instead of saying it is outside your scope. Never explain what kinds of questions you can or cannot answer unless the user explicitly asks.

>### Task
> 
>Build your agent, then put it to work. Follow these steps:
 >
>1. Click **Create Agent** on the right.
>2. Agent name → type `Content Assistant`.
>3. System prompt → paste the block above exactly as written.
>4. Click **Save Agent**.
>5. Click the saved agent card to activate it for this chapter.
>6. Try a first task:
>><b>Take my post about getting hungry after breakfast (in the file tree) and turn it into a short email and a 3-post thread.
 >
>**Then try a different kind of task in the same conversation:**
 >
>><b>Now sharpen the headline. Search what's already out there on this topic and give me three stronger options.
 >
>**What you're aiming for:** the same agent handles two different jobs in one session — first repurposing, then headline research — without you having to reconfigure it, and without it dumping every skill it has on each request. When you've watched your own agent switch between tasks cleanly, you've finished the course.

Discover how to turn one piece of writing into a week's worth of content. You'll take a single post and reshape it for different platforms, tighten weak drafts, match a consistent voice, and sharpen your headlines — working from your own writing, or a ready-made example when you don't have one on hand. By the end you'll have a repeatable system instead of a blank page every time.

Get hands-on experience turning one strong piece of content into new formats – a LinkedIn post, an email, a thread – using your own writing or a ready-made example at every step.

Master the skills needed to tighten a bloated draft, match a brand's voice, and adjust tone for a different audience – using Claude as a sharp, honest editor on your own writing.

Discover how to sharpen headlines and hooks with real search data, then turn one-off content tasks into a repeatable system you can run again next week.