6 Jobs to Give Claude Fable 5 Before July 7 — Full Prompts Included

Fable 5 is the most capable model Anthropic has shipped, and right now it's included in your Claude plan. On July 7 that changes — it moves to pay-as-you-go extra usage. So this is the week to stop asking it small questions and start giving it real jobs: the multi-step, run-for-an-hour kind of work it was actually built for.

Below are the 6 jobs I'd give it before the window closes. Each one comes with the exact prompt, copy-paste ready, plus the follow-up I use to push the result from good to done. They're written for Claude's agentic modes — Cowork or Claude Code — where it can read files, run code, and keep working without you babysitting it.

This is the fastest way to feel what Fable 5 can do. Take a screenshot of any landing page whose layout you like, drop it in, and let it rebuild the whole thing around your own brand. It's not tracing the design — it reverse-engineers the structure and rebuilds it clean.

The prompt: "Here's a screenshot of a landing page I like. Rebuild the layout from scratch with my own branding — colors, logo, copy — make it fully responsive, and keep refining it until it looks production-ready. Then give me a checklist of what to change before I deploy."

The follow-up: "Now audit it for accessibility and page speed, and fix everything you flag." This second pass is where it goes from a pretty page to something you'd actually ship.

Every developer has one: the old project with zero docs that everyone is afraid to open. Fable 5 will read the entire thing and hand you the map the last developer never made. This is the job where the bigger context window and longer autonomous runs earn their keep.

The prompt: "Read this entire codebase and explain how it actually works: map the architecture, list every module and what it does, flag dead code and risky spots, and write the README the last developer never did."

The follow-up: "Draw me a diagram of the data flow and mark the 3 files I should refactor first." You'll walk into that codebase on Monday knowing more than the person who wrote it.

Research is the most underrated agent job because it runs while you sleep. Give it your top competitors and let it dig through pricing pages, positioning, and what real users complain about in reviews. What comes back is the kind of report an agency would bill four figures for.

The prompt: "Research my top 3 competitors: their pricing, positioning, and what their users complain about most in reviews. Give me a report with the 5 gaps I can attack first, ranked by effort vs impact."

The follow-up: "Turn gap #1 into a landing page outline I can build this week." Research that doesn't turn into something shipped is just reading.

If you make content, this one pays for the whole week. One transcript in, a full content calendar out — and the key line in the prompt is 'study my last posts first', because that's what keeps it sounding like you instead of a generic AI feed.

The prompt: "Take the transcript of my latest video and turn it into a week of content: 3 short-form scripts with hooks, 5 posts, and a newsletter. All in my voice — study my last posts first. No recycled lines between pieces."

The follow-up: "Write 3 hook variations for each script and tell me which one you'd bet on and why." Making it argue for its picks is how you learn what it knows about hooks.

The least glamorous job on this list and the one that saves the most real hours. Point it at the folder of invoices and receipts you've been avoiding and let it do the extraction, the flagging, and the awkward follow-up emails in one run.

The prompt: "Go through this folder of invoices and receipts, pull every amount, date and client into a clean spreadsheet, flag anything unusual or unpaid, and draft the follow-up emails I still owe."

The follow-up: "Set this up so I can rerun it monthly with one prompt." That last line is the difference between a one-off trick and a system you keep forever.

The sleeper hit. Before your next pitch or negotiation, make Fable 5 play the other side of the table — and tell it to be difficult. It will push back harder than most real clients, and the scoring at the end is uncomfortably accurate.

The prompt: "Play my toughest client. I'm pitching [project] at [price]. Push back on everything — budget, timeline, scope. Negotiate hard, then score my answers 1–10 and tell me exactly where I folded and what I should have said."

The follow-up: "Run it again, but this time you're a client who ghosts and comes back with a lower offer." Rehearse the scenario you're dreading, not the one you're hoping for.

Don't try to run all six today. Pick the one that maps to something already on your plate this week — a real codebase, a real pitch, a real folder of receipts — and give Fable 5 the whole job, not a piece of it. The model rewards scope: the bigger and more clearly defined the task, the more the long autonomous runs pay off.

And when a run goes well, save the prompt with your tweaks. The window on free Fable 5 closes July 7, but a prompt library that fits how you actually work is yours forever — and every one of these jobs still runs on whatever model your plan includes next.