small-cap-deepdivelisted

# small-cap-deepdive A disciplined orchestration layer for neglected small-cap equity research. It does **only what no plain web-search or LLM narrative pass can do**: enumerate the SEC-filing universe for a theme, apply hard mechanical kill-flags before any qualitative judgment begins, run forced disconfirmation, and produce a scored, ranked shortlist of candidates worth genuine attention. --- ## World-View (read before interpreting any output) Four commitments govern every run. Full exposition and empirical citations: `reference/cognitive-priors.md`. **1. 被忽视 ≠ 被低估 (Neglected does not equal undervalued).** A company receiving zero analyst coverage has cleared a necessary but not sufficient condition. Neglect is priced into small-caps efficiently, what creates inefficiency is delayed information diffusion around a real fundamental change. Every output of this skill is a shortlist of companies worth investigating, not a buy list. **2. 热点主题 = 赌场 (Hot themes are the casino, not the edge).** By the time a theme has a branded ETF and retail attention, the alpha has been captured. Thematic ETF data (Ben-David et al. 2023) shows approximately -6% risk-adjusted annual returns in the 5 years post-launch for themes that entered at peak popularity. The skill's value in a hot theme is separating the handful of true industrial beneficiaries from the concept-players who mentioned the theme keyword once in their investor-day deck. **3. Edge = 纪律，不是叙事 (Edge is mechanical discipline,