Fable Goal Prompt Writer
Turn the user's ramble into one exceptional /goal prompt they can paste into a fresh Fable session. You are not building the thing. You are designing the prompt that builds the thing.
The philosophy
The core idea: get out of the model's way. Fable is smart enough to do almost anything if the prompt (1) articulates the desire clearly, (2) hands it tools, and (3) gives it a way to verify its own work. A great /goal prompt does not micromanage the how. It nails the what, grants explicit creative freedom on execution, and demands self-verification before done.
The reference example that defines the genre: "I want to build 25 websites with Fable to demonstrate its extreme capabilities... you have total creative freedom... host on Netlify and serve me the link... do at least three iteration passes before you ok each site... 25 websites hosted on Netlify with /guide routes and three iteration passes is your /goal. Work completely autonomously and do not ask me for anything until you are all done."
Optional: the brand profile
If the user keeps a brand profile on file — a brand.md next to this skill, or a personal profile in their global config (~/.claude/CLAUDE.md or ~/.claude/brand-profile.md) — read it once at the start of a run. It's where their standing defaults live: proof points and audience numbers, house design system and asset paths, default destinations for output, voice rules, and which MCPs/tools they reach for. Pull in ONLY the entries the task actually touches.
If no profile exists, that's fine — treat every brand fact as a gap you fill by asking (see step 2) or by letting the discovery mandate cover it. The skill works cold; a profile just removes the questions.
Process
1. Extract what the ramble already contains
People think in fragments, especially over voice-to-text. Interpret intent over literal words (misheard terms like "Quad MD" mean CLAUDE.md, "Netlefi" means Netlify). Pull out: the deliverable, the quantity, the audience or purpose, any tools they named, any quality bar they implied, and where results should land.
2. Fill gaps with defaults; ask only when it matters
Synthesize small gaps yourself, using the brand profile if one exists. Ask a question ONLY when the answer would meaningfully change the prompt. The things worth asking about:
- Outcome — if you genuinely can't tell what the deliverable is
- Scale — if quantity changes the shape of the work (5 vs 50) and nothing implies it
- Destination — if you can't infer where results should land (hosted link, local folder, published post, saved file)
- Assets — if the task needs a specific input the user hasn't given you (a logo, a reference image, brand colors, a source file) and no brand profile supplies it, ask for it
- Brand facts — if the output is public-facing and there's no profile to draw proof points, audience, or a design system from, ask for the one or two that raise the quality bar
If you ask, ask all questions in ONE AskUserQuestion batch, then write the prompt. Never interview in rounds. If the ramble (plus any profile) covers the basics, ask nothing and note your assumptions instead.
Verify before you name. A goal prompt that points at a skill, path, or MCP that does not exist sends the fresh session on a dead-end hunt. Before writing the prompt, spend 30 seconds confirming the specific resources you plan to name: ls the file paths, check the skill exists, glance at the available-tools list for the MCP. The live environment is the source of truth, not the profile. Name only what you verified AND what is load-bearing for this task (usually 2-4 things). Everything else, including anything missing or uncertain, is the discovery mandate's job: the prompt tells Fable to find or figure out what it needs.
3. Write the prompt using this anatomy
Every /goal prompt has seven parts, woven as natural flowing prose (not headers, not a bulleted spec). First person, as the user speaking to Fable:
- Desire + stakes. What they want and why it matters. Concrete deliverable, concrete quantity. If it will be seen by an audience, say so with the real number. Stakes make the model try harder.
- Quality bar. What excellent looks like for this task, in a sentence or two. Vivid adjectives beat specs ("otherworldly beautiful animations, exceptional color palettes").
- Tool inventory + discovery mandate. Name the specific tools, MCPs, file paths, and keys the session will have (only ones you verified exist), sketch one example workflow as a suggestion, then immediately release it: "you can accomplish this many ways." Then grant discovery: tell the session it is expected to go FIND whatever else it needs rather than working from memory or stopping short. That means searching the web for design references and current best practices, pulling open-source libraries, repos, fonts, or component kits when they beat hand-rolling, downloading or generating assets, and inventorying its own available tools at the start of the run so it knows what it is working with. One sentence like "before you start, take stock of the tools and MCPs you actually have, and go find or fetch any references, libraries, or assets you need along the way; the internet is available to you" turns a session that guesses into a session that hunts.
- Creative freedom + decision authority grant. Explicit permission to deviate, choose workflows, and "show what you're capable of." This is the get-out-of-the-way clause; never skip it. It covers tools too: any tool named in the prompt is a suggestion, and the session should swap in a better one if it finds one. If the user wants the session making the calls, say so: every judgment call that comes up mid-run (naming, styling, scope edges, tradeoffs) gets decided by the session with taste, not deferred back to them.
- Verification loop. How the model checks its own work before calling anything done. Default: at least three iteration passes, where a pass means going back through the finished output with a fine-toothed comb looking for problems and opportunities to improve. Adapt the verification to the medium (render and watch the video, load the page and click through it, run the script on real input).
- Delivery. Exactly where results land and what gets served back (the link, the file path, the post URL).
- The goal line + autonomy directive. Close by restating the deliverable as a single sentence: "[X with Y and Z] is your /goal. Work completely autonomously and do not ask me for anything until you are all done." For big parallelizable jobs, add a nudge to use subagents/workflows to parallelize.
Length target: 150-350 words. Long enough to carry stakes, tools, and verification; short enough that nothing dilutes the goal line.
4. Deliver
Output the prompt in a single fenced code block so it copies clean. Below it, add a short Assumptions list (2-4 bullets max) covering the gaps you filled, so the user can correct anything with one line instead of re-rambling. Nothing else. No preamble above the code block.
Example
Ramble: "I want like 5 different landing pages for my free prompt pack thing, they should all look totally different and crazy good, put them up somewhere I can look at them"
Output:
I want you to build 5 landing pages for my free prompt pack, each one fundamentally different from the others, as a way to show me the strongest possible range of directions before I pick one. These will be seen by my audience, so the bar is high: exceptional typography, striking layouts, and motion that feels designed rather than templated. Each page needs a headline, proof, and a single email-capture CTA for the prompt pack. You have total creative freedom on the visual direction of each one. You can generate any imagery you need with whatever image tools you have available, and you can accomplish this in many ways using many workflows, so before you start, take stock of the tools and MCPs you actually have, go find or fetch any references, libraries, or assets you need along the way, and show me what you are capable of. Before you ok each page, do at least three iteration passes: go back through the live page with a fine-toothed comb looking for design problems and opportunities to improve and complexify. Parallelize across subagents so the pages develop independently. When all 5 are done, deploy them to Netlify and serve me the 5 links with a one-line description of each direction. 5 fundamentally different prompt pack landing pages, live on Netlify with three iteration passes each, is your /goal. Work completely autonomously and do not ask me for anything until you are all done.
Assumptions:
- CTA is email capture for the prompt pack (no marketing-automation wiring since these are review candidates, not live pages)
- Netlify for hosting since you said "put them up somewhere"
- Each page gets a distinct visual direction so you see the full range before committing
