Build the Vault

The tracker did the hard part: it ranked your niche by a virality score and floated the breakouts to the top. Now the question is what made those posts break through — and the answer is almost always two things: the hook that stopped the scroll, and the CTA that converted the attention.

The vault extracts both from the outliers the tracker already found, and stores them as two databases you can sort by proof.

Two databases, both ranked by proof

You're building two files:

1. hooks.json — the opening line of each outlier, tagged by type: bold claim, question, story tease, results reveal, contrarian, FOMO. Each one carries the outlier_multiplier of the post it came from.
2. ctas.json — how each outlier asks for the action: comment, follow, DM, keyword. Again, tagged with its multiplier.

The multiplier is the whole point. You're not collecting hooks — you're ranking them by how far the post they came from beat its creator's own median. A "contrarian" hook attached to a 12x outlier tells you something a "contrarian" hook attached to a 2x post doesn't.

The extractor prompt

Paste this into Claude Code. It reads the outliers straight out of the tracker's database and writes the two vault files.

Read the posts with outlier_multiplier >= 3 from track.db. For each one, extract and save into two files:
(1) vault/hooks.json — the opening line/hook, categorized by type (bold claim, question, story
    tease, results reveal, contrarian, FOMO), with its outlier_multiplier.
(2) vault/ctas.json — how it asks for the action (comment, follow, DM, keyword), with its multiplier.
Rank them by multiplier so I can see which hooks and CTAs actually work in my niche.

Notice the threshold: outlier_multiplier >= 3. Back in the tracker chapter, 2x means "worth noticing" but 3–5x is "strong." The vault only cares about the strong ones — there's no point learning from posts that barely beat their own average.

A living swipe file, not a dead screenshot folder

Most people's "swipe file" is a camera roll full of screenshots. It's a graveyard: you can't sort it, you can't tell which hook actually worked, and by the time you need it you've forgotten why you saved each one.

The vault is different because it's built from data and ranked by performance:

• It lives in files, so it grows every time you re-run the tracker.
• Every entry is stamped with proof (the multiplier), so you sort by what worked in your niche.
• Because data/mine/ and data/refs/ share one schema, the vault can eventually rank your hooks next to your competitors' — the same swipe file, cross-referenced.

That's the difference between a swipe file that compounds and one that just fills up.

Key takeaway

The vault reads the tracker's outliers (multiplier >= 3) and extracts two ranked databases — hooks tagged by type, CTAs tagged by ask — each stamped with its virality multiplier. It's a living swipe file ranked by proof, not a dead folder of screenshots.