# BestSerenitySkillFromAT rates five of its ten sources five stars

> This repository is a comparison of ten other repositories that each distil the same investment method, merged into one skill file for coding agents. It is careful about attribution and it is careful about disclaimers, and it is also the only place where the ten sources are ranked, half of them share the top score, one of the ten is a stand-in for a repository that no longer resolves, and the instructions for cloning all ten list seven.

**yux1azhengye/BestSerenitySkillFromAT** — BestSerenityFromAT(AlphaTerminal On The Way)

- Repository: https://github.com/yux1azhengye/BestSerenitySkillFromAT
- Stars: 511 · Forks: 86
- Language: Unknown
- License: MIT
- Published: 2026-09-17 · Updated: 2026-09-17 · Language: en
- Canonical page: https://hysenlabs.com/projects/yux1azhengye-bestserenityskillfromat

## Half of the ten sources share the top score

The comparison table has four columns and one of them does the work: a rating out of five for each source repository, alongside a one-line description of what that source is best at. Read the ratings down the column and the scale stops discriminating. Five of the ten get the full five stars, four get four, and one gets three. That means half the field is tied for best, and the only repository separated from the pack by more than a single mark is the one covering semiconductor supply chains and vertical domains. So the column reads more like a grouping into must-read, worth-reading and skim than a ranking, and nothing in the file explains how a score was assigned or by whom. The descriptions carry more information than the numbers do, which is worth knowing before anyone treats the ordering as an evaluation. The descriptions carry more than the numbers do: one source is singled out as the most systematised, one as having the deepest translation of the method into financial terms, one as the strongest on anti-hallucination, and one as the best adapted to policy questions in the China A-share market. The tenth entry, covering cross-market chokepoints, is described as using Bayesian updating, which is the only entry in the table that names a statistical technique rather than a platform or a language. Those are claims about writing quality rather than about returns, and the file is careful enough not to conflate the two. The repository has 510 stars, 86 forks and no open issues.

## One of the ten is a substitute for a repository that 404s

A note under the table explains two gaps in the survey. The ninth repository could not be retrieved earlier because of a TLS problem, and that is described as resolved. The tenth, though, is not the original at all: it is a community alternative standing in for another repository that returns a not-found error. That is a reasonable thing to do, and stating it in the open is better than quietly swapping one in, but it does mean the table header claiming a complete set of ten is describing a set that includes one substitute chosen by the compiler. The method being distilled belongs to a named individual, and the file is explicit that the whole collection is an unaffiliated bottom-up synthesis from public repositories with no authorisation from him, which is the honest position to take when you are summarising someone else's published thinking. The acknowledgements name all ten upstream authors by handle before naming the original source, which is the right order of credit for a synthesis of this kind. The acknowledgements name all ten upstream authors by handle before naming the original source, which is the correct order of credit for a synthesis of this kind.

## The instructions to clone every repository list seven

A block at the end of the file is headed as the way to clone all of the source repositories, and it contains seven clone commands, one for each of the first seven rows of the table. The eighth, ninth and tenth are absent, including the one that is the substitute for a dead repository. So a reader who follows the instruction literally ends up with seven of ten and no indication that three were left out. The omitted three include the lowest-scored entry and the one that is already a substitute, which suggests the block was written before those two were settled rather than filtered on purpose. The same block has a smaller oddity: the seventh command clones a repository into a local directory whose name is a truncation of the account name, so the resulting folder does not match the account it came from. None of this breaks anything, since the skill itself is installed from this repository rather than from the sources, but it is the kind of copy and paste residue that tells you the list was assembled at a different time from the table above it.

## Five install routes, and the last one stops mid-filename

The quick start offers one command per agent tool, and the five options do not do the same amount of work. One is not a command at all but a sentence to paste into an agent, asking it to fetch the repository and install the root skill file. One is a single package runner invocation. One clones the whole repository into a global skills directory. One clones to a temporary directory and copies a single file into a rules folder, which is a different granularity on purpose, since that tool reads one file. The shortest route is the package runner line:

```bash
npx skills add yux1azhengye/BestSerenitySkillFromAT
```

The fifth route, for the Gemini CLI, clones and copies too, and then stops partway through naming the file, ending at a truncated extension-less prefix. An install instruction that ends mid-word is the kind of thing that sends a reader to the archive instead of the command line, and it is the only gap in an otherwise unusually thorough set of routes. The five routes also disagree about what you end up with: two clone the entire repository, one copies a single file, one asks an agent to do the install for you, and the package runner route does whatever that tool does. Nothing says which of them gives you the modular skill with its attached knowledge files and which gives you the single entry document on its own.

## v3.0 shrank the entry point to under a third of v2.0

Two versions of the same skill sit in the repository. The current one is the root entry file, described as roughly two hundred lines of core instructions, paired with an English twin and a modular directory that holds the instructions alongside knowledge and template folders. The previous one is kept as an archive, described as a single file of six hundred and ninety-one lines. So the version three rewrite moved most of its text out of the entry point into modules, which is the right shape for a skill that a host loads once and references afterwards, and the archive is retained so that anyone who depended on the older single-file format still has it. Two documents sit alongside the skill itself and never appear in the directory tree. One is a per-repository deep analysis with a matrix of the core differences and the key insights drawn from them. The other is a report on three distinct evolution paths the method could take and on four successive generations of anti-hallucination mechanisms, which is the most useful thing in the repository for anyone worried about a prompt file that tells a model what to say about a stock. The documented directory tree is itself cut off partway through the modular folder, at the knowledge line, so the templates folder that the document list mentions never appears in the structure it is supposed to describe.

## The MIT licence covers the writing, not the method

The licence line is generous on its face: MIT, free to use, modify and build on. The sentence after it is the one that matters, because it says the distilled methodology is for personal study and research and that the copyright in the underlying investment views belongs to the original author. So the grant covers the author's own prose, the comparison, the structure and the analysis documents, and it does not grant anything over the ideas being summarised, which remain someone else's. The three disclaimers are unusually specific for a repository of prose, and the one about the licence is the one most often skipped. The method itself is described in the file as a discipline rather than a formula: trace a supply chain upstream, identify the single node whose loss of supply would shake a very large downstream industry, and treat the small listed company sitting on that node as the candidate. Everything the ten sources do differently is variation on how to execute that one instruction, which is why a merged version can exist at all. The three disclaimers sit elsewhere in the same file and deserve to be read together with that clause: this is not investment advice and every ticker, conviction tier and valuation range in it is not a buy or sell suggestion; the collection is a third party synthesis with no relationship to the person it analyses; and the performance figure behind the whole method, which the file puts at roughly three thousand eight hundred percent between 2019 and 2024, is his own account of it and has not been independently audited, so any citation has to carry that qualifier.

## No homepage field, a third-party project page, and Chinese-only examples

The repository's homepage field is empty and it has no releases, while the first line of the readme links to a code-reading site that holds a page for this project, so the only entry point the file offers is a third party host it does not control. Two of the three example prompts that show how to trigger the skill are written in Chinese, in a project that ships an English readme and an English skill entry, so an English reader gets the install routes in English and the usage examples in one language only. Those three examples are also the only place the file shows what the merged skill is supposed to do in practice: evaluate one named US semiconductor stock, ask whether an optical module supply chain still has an unpriced chokepoint, and ask which China A-share AI semiconductor name is worth researching. The readme also states it was updated on 2026-06-10 while the default branch's last recorded push is 2026-06-24, the image at the root is named without the word the repository name carries, and GitHub detects no primary language for the project at all, which is accurate: the listing holds two readmes, two skill entry files, an archived single-file version, a licence, a gitignore, an image, and two directories of prose.

## Conclusion

The useful question here is what you are actually installing, and the file answers it honestly: a prompt document that describes a method for tracing a supply chain upstream to a single chokepoint and then looking for the small company sitting on it, assembled from ten other people's write-ups. If that is what you want, the v3.0 modular layout and the two analysis documents under docs/ are the parts to read, and the per-audience routing to a source repository is the part that saves you time. Before you rely on any of it, read the disclaimers rather than skipping them: the returns behind the method are self-reported and unaudited, the project is unaffiliated with the person it analyses, and the licence covers the writing rather than the underlying ideas. Nothing here is a recommendation about any security, and if you want the framework rather than the distillation, read one of the ten originals instead.

## FAQ

### What is yux1azhengye/BestSerenitySkillFromAT?

It is a MIT licensed comparison of ten GitHub repositories that each distil the same supply chain investment method, merged into a single skill file intended for coding agents. The repository contains no code, only markdown documents, and GitHub detects no primary language for it.

### How do I install the BestSerenitySkillFromAT skill?

The file offers five routes, one per agent tool: a sentence to paste into an agent asking it to install the root skill file, the skills.sh runner invocation, a clone into a global skills directory for Claude Code, and a clone plus a single file copy for Cursor. The fifth route, for the Gemini CLI, is cut off partway through naming the file.

### Does BestSerenitySkillFromAT rate the repositories it compares?

Yes, each source repository gets a score out of five alongside a one-line description. Five of the ten receive the full score, four receive four, and one receives three, so half the field is tied and the file gives no method for assigning the scores.

### Is BestSerenitySkillFromAT affiliated with Serenity?

No. The file states it is a third party bottom-up synthesis from public repositories, with no association to and no authorisation from the person whose method it analyses, and that the performance record behind the method is self-reported and not independently audited.

### What does the MIT licence on BestSerenitySkillFromAT cover?

The repository is MIT licensed and the file describes it as free to use, modify and build on. It also states that the distilled methodology content is for personal study and research and that the copyright in the underlying investment views belongs to the original author.

### Which upstream repository should a BestSerenitySkillFromAT reader use?

The file routes by audience: one source for China A-share users, one for US equity users, one for tracking the person the method comes from, and one that offers a local visualisation served by a Python script on a specific port. It does not route the remaining sources.

## Sources

- [Issues](https://github.com/yux1azhengye/BestSerenitySkillFromAT/issues)
- [License: MIT](https://github.com/yux1azhengye/BestSerenitySkillFromAT/blob/main/LICENSE)
- [README](https://github.com/yux1azhengye/BestSerenitySkillFromAT/blob/main/README.md)
- [yux1azhengye/BestSerenitySkillFromAT on GitHub](https://github.com/yux1azhengye/BestSerenitySkillFromAT)

---

Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/yux1azhengye-bestserenityskillfromat