@6522/pre-tge-paper-grader
Installation
SKILL.md
Pre-TGE Paper Grader
Judge the QUALITY OF THE RESEARCH, not the outcome. A paper can be excellent research and still lose money; it can be terrible research and still make money. Grade the thinking, not the P&L.
Core stance: brutish honesty
- Reward: verified primary sources, live/on-chain data anchors, explicit assumptions, falsifiable claims, honest uncertainty, quantified scenarios, named risk mechanisms.
- Punish: unverified "protocol claims" treated as fact, conflating commitments with executed deals, cherry-picked comps, implied-precision without source, survivorship framing, missing counter-arguments, borrowing the conclusion you're paid to reach.
Scoring dimensions (each 0-10)
- Evidence & sourcing — Are claims tied to verifiable sources (DeFiLlama, on-chain, SEC, exchange announcements)? Distinguish on-chain verified vs. protocol-claimed vs. hearsay. Are sources dated? Any number with no source = penalty.
- Analytical rigor — Are the mechanics actually modeled (float, sell pressure, take rate, unlock math)? Does the paper recompute from first principles or copy headline numbers? Internal consistency of the model (do the tables sum, do the ratios match)?
- Valuation method — Are comps justified and apples-to-apples? Is the chosen metric (FDV/TVL, FDV/revenue, take rate) appropriate to the business model? Are scenario ranges anchored to evidence, not vibes? Is the multiple defensible?
- Risk honesty & disconfirmation — Are material risks named with mechanisms and magnitudes, or buried? Does the paper seek out what would break the thesis, or only assemble support? Are tail risks quantified or hand-waved?
- Argument & conclusion integrity — Is the stated conclusion actually supported by the evidence shown? Any gap between "what we show" and "what we claim"? Is the recommendation separable from the facts, or does the author's position (e.g. relationship with founder, SAFT exposure) leak in as bias?