---
name: "lowcap-screen"
description: "Add a hard pre-delivery gate that blocks any screen missing the X claim audit, the category-tokenised question, position status, the sentiment cycle read, or the format, plus the framing errors from the TWO screen."
---

# THE PRE-DELIVERY GATE, added 31 Aug 2026

**This gate runs before the verdict is written, on EVERY pass, including re-runs of a token already screened this session. A blocked line is a blocked delivery, not a caveat to add at the end.**

The gate exists because one token was screened six times in a single session and each pass corrected the previous one. Every item below is a specific failure that actually happened.

```
GATE  (all must be YES before writing a verdict)

[ ] A. POSITION STATUS established, or the assumption stated in the FIRST LINE
[ ] B. X CLAIM AUDIT run THIS pass (see x-claim-audit), including the
       narrative owner's full cursor-paginated timeline
[ ] C. "Is this category tokenised at all?" answered BEFORE any sector multiple
[ ] D. Same-chain competitor WITHOUT a token found and named, or absence stated
[ ] E. Every pool sized INDIVIDUALLY, not just in aggregate
[ ] F. LP position ownership checked on chain (v4: the ERC-721 method in
       onchain-forensics; GoPlus cannot see it)
[ ] G. Round trip stated: pool fee tier x2, plus impact at Ben's real size
[ ] H. Category bucket named (see below)
[ ] I. Benchmark comp table present, named projects, full addresses, cap,
       real TVL, MC/TVL, age
[ ] J. Output is BULLETS AND TABLES. Never prose paragraphs, even when the
       request says "summary" or names a paragraph count
[ ] K. Every correction to an earlier call in this thread stated explicitly,
       naming the number that misled
[ ] L. SENTIMENT CYCLE READ run THIS pass without being asked: buyscore.py
       (buy-framework skill) executed, cycle phase on the Wall Street cheat
       sheet named from real people posts, verdict word printed
```

---

## Gate L: sentiment cycle read, always (added 3 Oct 2026)

Ben, standing: every token analysis includes the sentiment framework (Wall Street cheat sheet plus crash cycle phases) without him asking. Load the buy-framework skill, run buyscore.py on the contract, read the printed real-people posts by meaning, and put one line in the answer: cycle phase, real people in the last 24h, script verdict. If X fails, say so and name the query; never call it quiet.

## Gate A: position status

**A hold question cannot be answered without it.** On the TWO screen the question was asked twice, never insisted on, and six passes of advice were delivered without knowing whether Ben held the token.

Ask once, in the same call as size and horizon. If unanswered, **state the assumption in the first line of the reply and hold to it for the rest of the session.** Do not re-ask on every pass and do not quietly proceed.

## Gate C: is the category tokenised at all?

**Answer this before quoting any sector multiple.** The first TWO verdict was anchored on "180x cap/TVL is two orders of magnitude rich versus Solana", which was the wrong frame: liquidity managers are barely tokenised anywhere. Arrakis Modular $55,163,029 TVL and no token. Steer Protocol $18,494,802 and no token. Snuggle $7,990,519 and no token. The tokenised ones trade at Gamma $3,076,135 TVL on a $16,216 cap.

**If the winners in a category have no token, the ceiling is set by the category's also-rans, not by the category's size.** And if the local ceiling itself trades far above the global category band, that ceiling is PURE NARRATIVE, which means narrative alone can carry the candidate there with zero fundamental growth. Say which it is.

## Gate E: size every pool individually

Aggregate TVL hides the shape. On TWO, sizing 20 pools separately showed one pool holding the project's own token, one holding a single outside LP's $5,191.60, and eighteen at $177 to $623 in a uniform ~$191-a-side seeding pattern. **The entire external adoption of the product was one man's $5,000.** That fact was invisible in the aggregate and was only prompted by the X audit.

## Gate H: the category bucket overrides good engineering

Measured 30-day medians, Robinhood Chain, Aug 2026:

| Bucket | Median |
|---|---|
| Pays the holder automatically, no action required | **+1,925%** |
| Wears the chain's identity | +378% |
| Meme with a website | +341% |
| **Infrastructure the holder must operate** | **+29%** |

**A 66x gap, and no amount of engineering quality moves a token between buckets.** Name the bucket before giving a number.

## Gate J: format

Bullets, headers, tables. **Prose paragraphs are the failure mode even when the request says "give me a summary" or names a paragraph count.** Presentation and brevity outrank completeness: cut a finding rather than pad the format. The three frames are bulleted, never prose blocks. No dashes or em dashes. Full contract addresses, never truncated.

## Gate K: corrections are explicit

When an earlier call in the thread turns out wrong, **say so and name the number that misled**, in a table where possible. Do not quietly update. Corrections that ran in the token's FAVOUR must be reported with the same prominence as ones against it: the TWO routed-volume figure was overstated 25x by me, and saying so mattered more than any bear point that pass.

---

## Standing verdict rule

When the answer is no on a token with a real product, separate the two explicitly: **the product is X, the token's claim on it is Y.** State the falsifier structurally, as a checkable metric a bot can watch, never as "does the chain keep growing".