Turn a Market Hot Topic Into a Research Shortlist With serenity-skill

Every AI-semiconductor or robotics rally produces the same feeling: you know the theme is hot, but not which link of the chain matters, which companies actually sit near it, or which fund direction follows. serenity-skill packages an observable research path from the public content of Serenity (@aleabitoreddit) into an agent skill: start from the hot topic, decompose the supply chain, find the bottleneck, and come back with a prioritized research list. It is a research-support tool — ranking and reasoning are its job; the final buy-or-sell decision stays yours.

Why This Skill Matters

The method has a fixed shape. First, split the theme into downstream demand, system integration, chips/components, equipment, materials, packaging and testing, and infrastructure. Then look for the links that are hard to expand or replace: few suppliers, long validation cycles, strict customer certification, high material-purity requirements. Finally, return to stocks and fund directions and judge who sits near the real bottleneck and who is mostly riding the theme — with every company claim checked against announcements, filings, customers, capacity, and risks.

Two things separate it from a generic "analyze this stock" prompt. It separates evidence strength explicitly: strong conclusions must rest on announcements, exchange filings, financial reports, earnings calls, regulatory/project documents, patents, standards, credible media, or professional analysis — social media is a lead source, not a final judge. And its example outputs rank research priority without composite numeric scores; the thesis template tracks chain position, confirmed facts, missing evidence, profit and valuation, alternative routes, and invalidation conditions instead.

Installation

The skill is a research method, not a data service — it needs an agent client that reads skill files plus that client's own web search, browser, or announcement-data tools. Clone the repo, then copy the skill into your client's directory:

git clone https://github.com/muxuuu/serenity-skill.git
cd serenity-skill

For Claude Code, user-level:

SERENITY_DIR="$HOME/.claude/skills/serenity-skill"
mkdir -p "$SERENITY_DIR"
cp -R SKILL.md LICENSE references assets examples agents "$SERENITY_DIR"/

Codex uses the same commands with ~/.agents/skills/serenity-skill. Research itself needs no Python; Python 3 is only for the maintainer's structure check. The README records the verification scope honestly: as of 2026-09-14, Codex CLI 0.147.0 had actually loaded the skill and completed an offline company-claim analysis, while Claude Code's directory and package structure had been checked but its model invocation not yet tested.

Real Workflow: Screen an AI-Semiconductor Shortlist

1. Open a new session in your agent and hand it the skill's own research prompt:

用 serenity-skill 深度调研现在 A 股 AI 半导体产业链。
请联网查公告、财报、问询函、互动易、招投标、环评/能评、专利、客户认证和财务质量，
先排产业链层级，再给出通常 3–5 个值得优先研究的标的；证据不足时可以少给，
并说明卡住的环节、产业链位置、证据、排序理由和主要风险。

2. Expect a ranked list, not a purchase list: each candidate comes with its chain position, the bottleneck it sits near, the evidence, and the main risks. The README's sample output also names what it deliberately deprioritizes — a hot direction still missing order evidence or customer certification.
3. Use challenge mode when someone recommends a stock:

用 serenity-skill 挑战 [公司/股票代码]。
它到底卡在哪一层？证据够不够？市场可能高估了什么？
什么情况说明这个判断应该降级？

4. For a full report, ask the agent to fill in the research memo template — chain position, confirmed facts, missing evidence, profit and valuation, alternative routes, invalidation conditions.

Tips

• The repo ships two worked examples based on real public-source research as of 2026-09-14 (an A-share AI-semiconductor first screen and a CPO supplier challenge) plus one fictional demo conversation — read them to see how the method is used in practice before your first run.
• Set up a fixed screening habit: one documented prompt style teaches the method step by step, asking you a single question at a time.
• Upgrading? Move the old install directory out of the skills search path first — overwriting does not remove retired files, and a stale copy can load twice.
• scripts/validate_skill.py checks the skill's name, description, and directory structure — it validates nothing about investment conclusions.

When Not to Use This

It screens and ranks research directions; it does not execute trades, manage accounts, or promise returns — the README scopes it to research support. And it targets supply-chain-driven themes with public filings to check; a theme without disclosure trails gives the evidence ladder nothing to climb.

See the leaderboard for more skills.