> For the complete documentation index, see [llms.txt](/llms.txt).

Every listing that reaches your results feed gets a score out of 10, an action (ALERT, WATCH, or SKIP), and a one-sentence verdict explaining the decision. Here's what's behind each.

## The verdict thresholds

- **ALERT (score 7+)** — matches your criteria and the price is fair. This is "tell a friend about it."
- **WATCH (score 5–6)** — close, but something gives pause: pricing near your ceiling, a low-feedback seller, or a minor spec mismatch.
- **SKIP (score 1–4)** — wrong product, scam signals, overpriced, or missing critical info.

Nothing is skipped silently. A SKIP lands in your results feed with the verdict so you can see *why* it missed — usually a criteria or pricing problem you can fix.

## The seller-risk rule

One rule overrides the score: a listing from a seller with **zero feedback, a brand-new account, or a rating below ~95%** can never be an ALERT. The seller risk caps it at WATCH no matter how good the price looks. If a deal feels too good to be true and the seller is anonymous, that's the system doing its job.

## What's checked before scoring

Before a listing is even scored, it passes through a filter stack. A listing is dropped early (as a SKIP) when it:

- Doesn't match the hunt's category
- Fails keyword or source rules
- Priced above your ceiling, or has a broken/missing price
- Is a duplicate you've already seen
- Scores below the relevance floor (a weak match is discarded before the full scoring pass)

Bundles and lots aren't hard-rejected — they're flagged so you can judge the math yourself.

## What the score reads

The scorer looks at the listing title, price, description, and your hunt's context — brand, capacity, and condition tolerance — and checks three things:

1. **Is it the right product?** Right spec, right generation, right capacity.
2. **Are there red flags?** Wrong spec, vague listing, new seller, implausible price.
3. **Is the price fair?** Anything above your ceiling was already filtered out before scoring — the score weighs whether the price is reasonable for what you're getting.

The verdict is one plain-language sentence: *"Matches spec, $60 under the median, established seller."* Read the verdict, not just the score — it's why the number is what it is.

## The price facts

Your **ceiling** is the only price number the filter uses — anything above it is dropped. Your **target** is your own act-fast reference; it doesn't change the score. If you want more (or fewer) results, adjust the ceiling and the criteria, then let the next scan re-evaluate.