The 5-Prompt Content Engine (The Rebuilt Version of Every Viral Prompt Pack)

You've seen the prompt packs: six copy-paste prompts, "watch your account grow." I've written before about why they underdeliver: a borrowed prompt knows nothing about you, so it fills the gaps with the average of the internet, and you get back everyone's content instead of yours.

But the SEQUENCE in those packs is usually right: understand your audience, find ideas, package them, repurpose what works. So here's my rebuilt version. Same bones, five prompts, with one difference that changes everything: every prompt is wired to YOUR files. Your brand notes, your voice rules, your shipped content, your numbers.

What you need before running these

The engine reads three things. Thirty minutes of setup if you don't have them:

1. A brand file: who you are, what you sell, who it's for. A page is plenty.
2. Your voice rules: how you write, what you never say. (The rules doc guide builds this in 20 minutes, out loud.)
3. Your shipped content: the last two months of what you've actually posted, in a folder your AI can read, ideally with a note on what performed.

Point your AI at those (a project, a folder, or a second brain if you've built one) and the prompts below stop guessing.

Prompt 1: the audience map (run once, revisit quarterly)

The generic packs ask the AI to imagine your audience. This one forbids that.

Read my brand file, my voice rules, and my last two months of shipped
content before answering. Do NOT invent my audience. Build the map
from what those files actually say, plus what the shipped content
implies is landing.

For my audience, map:
- Their 5 biggest frustrations, in their own likely words
- Their goals: what "winning" looks like to them this quarter
- Their fears and objections about [my topic/offer]
- The outcome they'd actually pay for
- The questions they're Googling and asking friends

For each, name the ONE content format most likely to earn a watch, a
save, a share, a comment, and a follow, and say why. Flag anything my
files don't support as an assumption to confirm, not a fact.

Save the output as a file. Every prompt below reads it.

Prompt 2: the idea finder (weekly)

The generic version asks for fifty ideas and hands you six usable ones. This one optimizes for fit.

Read my audience map and scan my shipped content so you don't repeat
anything from the last 8 weeks.

Generate 15 content ideas. Draw from: my audience's mistakes, myths
they believe, lessons I've learned, contrarian takes I can back up,
and "how do you actually do X" moments from my real work.

For EACH idea give me:
- One-line concept
- Lane: REACH (widely relatable, emotion-first) or TRUST (tool demo,
  how-to, build-along)
- The emotion it stirs + the curiosity gap it opens. If you can't
  name both, cut the idea before it reaches me.

Then rank them, tell me the healthy weekly mix of reach vs. trust,
and pick the 3 to 4 you'd make first, with reasons. Quality over
volume: 15 real ones beats 50 generic ones.

Prompt 3: the packager (per idea you'll actually make)

Hooks in nobody's voice are the signature failure of borrowed prompts. This one is fenced by your rules.

Read my voice rules first. They are non-negotiable.

Content idea: [paste the idea]

Package it:
- 3 hook options in MY voice (no dramatic tee-ups, no "the secret?",
  nothing I'd have to perform). For each: the emotion + the curiosity
  gap it opens.
- Text-on-screen line for the first frame that frames the watch
- Structure in reverse: open (hook), hold (the one open loop that
  runs the whole piece), land (the payoff, withheld until the end)
- One CTA that fits the lane: soft save/follow for reach, a specific
  next step for trust

If the idea only works with manufactured drama or a bait hook, say
so and kill it instead of packaging it.

Prompt 4: the repurposer (after something ships and works)

The generic packs turn one post into ten near-clones "keeping the core message intact." That's the same post wearing ten hats, and audiences can tell.

Read this piece that already shipped and performed: [paste or point
to it]. Read my voice rules.

Repurpose it into [a carousel / a LinkedIn post / a newsletter
section / a short video script], with each version REBUILT for its
surface, not trimmed to fit:
- Different entry point per surface (what stops a scroll on IG is
  not what earns a click on LinkedIn)
- Same underlying idea, new specifics or angle per version, so
  nothing reads as a duplicate to someone who follows me everywhere
- Platform limits respected, in my voice throughout

If a surface doesn't suit this idea, say skip it and why.

Prompt 5: the orchestrator (when you want the week in one pass)

Once prompts 2 through 4 feel routine, chain them:

Run my weekly content pass:
1. Idea-find per my idea finder prompt (respecting everything
   already shipped)
2. For the top 3, package each per my packager prompt
3. Flag which shipped piece from the last two weeks most deserves
   repurposing, and to which surfaces

Deliver: the ranked idea list, 3 packaged concepts ready to film,
and the repurposing recommendation. Anything that fails my voice
rules or needs bait to work gets cut, not fixed.

One run, and your week is planned from your own material.

Why this beats the pack you screenshotted

Line the two up. The generic pack guesses your audience; this reads your files. It asks for volume; this screens for fit. Its hooks belong to nobody; these are fenced by your voice rules. Its repurposing clones; this rebuilds per surface. Same skeleton the whole way down, and every difference is the same single move: the AI stops guessing because you handed it what it was missing.

The prompt was never the valuable part. Your context is.

FAQ

Do these work in ChatGPT or only Claude? Any tool that can read your files. The engine's requirement is access to your brand file, voice rules, and shipped content, not a specific model.

What if I don't have shipped content yet? Run prompt 1 against your brand file and any writing you have (emails, a bio, notes). Mark more as assumption-to-confirm, ship your first month, then re-run with real data. The engine gets sharper every week it has more of you to read.

How long does the weekly pass take? With the files in place, prompt 5 runs the week in one sitting: 20 to 30 minutes including your review. The review is not optional; you're the editor, always.

Isn't this just another prompt pack? It's the anti-pack: five prompts whose whole design is refusing to run without your context. Copy them, but the work that makes them sing is the 30 minutes of setup, and that part can't be screenshotted.