Small Cap Deepdive — Development skill for Claude Code

small-cap-deepdive

Mechanically de-risk the SEC small-cap universe for a theme or ticker, kill the landmines, then deep-dive the survivors.

[](https://docs.anthropic.com/en/docs/claude-code) [](LICENSE) [](#-read-this-first--the-design-philosophy) [](https://github.com/dgunning/edgartools) [](#languages) [](ROADMAP.md)

[English](README.md) | [中文版](README_CN.md)

---

⭐ Read this first, the design philosophy

**Being neglected is not the same as being undervalued.**

Why that is true, and what it takes for a neglected name to also be mis-priced: [`reference/cognitive-priors.md`](reference/cognitive-priors.md) §1.

**The output is a landmine-scanner, not a buy list.**

What a top-ranked name does and does not entitle you to conclude: [`reference/cognitive-priors.md`](reference/cognitive-priors.md) §5.

**Zero buys is a feature, not a bug.**

If a theme produces no score-4+ candidates, the tool is telling you the small-cap universe for that theme does not contain clean industrial beneficiaries at this time. That is a correct and useful answer. A scanner that cannot say "nothing found" is not a scanner; it is a narrative generator.

The one-sentence version: **the tool's edge is mechanical discipline applied consistently across the full candidate set, not narrative synthesis on any individual company.** Every tool, invariant, and hard rule in this repo exists because of four principles, root-cause design (not symptom patching), Hybrid-not-thin (the data layer earns its keep), discipline-as-moat, an