AleaBito / Serenity Skills
A suite of Claude / Codex agent skills distilled from @aleabitoreddit ("Serenity")'s entire public archive — ~11 months, 6,120 posts, 750 tickers (2025-07-02 → 2026-05-30). Track her, analyze like her, and anticipate where her attention is moving — grounded in real data, not vibes.
📊 Live dashboard / 实时看板: Lifetime attention tracker · 全周期注意力看板 — built on top of these skills · 基于这套技能制作
claude-skill · agent-skill · aleabito · serenity · investing · stock-research · supply-chain · first-principles · x-api
English
What this is
Three interlocking skills that turn one prolific X (Twitter) research account into a reusable workflow:
Skill What it does Key command
follow-aleabito
Data layer. Fetches her posts via the X API, builds a beginner-friendly digest, cumulative ticker-mention analytics, and a durable research map. Includes a full-archive backfill.
node skills/follow-aleabito/scripts/analyze-mentions.js --incremental --resume
serenity-method
Her analysis method, generalized. Apply her style to any stock: critical-chokepoint / supply-chain-OSINT discovery → first principles → a Buffett-style quality gate (fields default to unverified) → narrative-vs-fundamentals hygiene → research-map vs investable-conclusion.
(invoked by request: "analyze $X like Serenity")
serenity-radar
Where her attention is going. Reads the mention archive for attention momentum (heating tickers, new entrants, conviction core, theme rotation) and generates candidates in her style — a candidate generator + checklist, never an oracle.
node skills/serenity-radar/scripts/radar.js --window 14
The pipeline: follow-aleabito (data) → serenity-method (how she reasons) → serenity-radar (what she's likely to focus on next). Each later skill builds on the earlier one.
Install
- Copy the skills into your agent's skills directory:
cp -R skills/* ~/.codex/skills/ # Codex
# or
cp -R skills/* ~/.claude/skills/ # Claude Code
-
Node 18+ is required for the scripts.
-
X API token — put a bearer token in
~/.follow-aleabito/.env:
X_BEARER_TOKEN=your_token_here
The full-archive backfill (--archive) needs an X API project entitled to full-archive search (/2/tweets/search/all); the daily incremental fetch works on the standard user-timeline endpoint.
- Point at your data dir (where the analytics CSVs live):
export FOLLOW_ALEABITO_REPORTS_DIR="$HOME/aleabito-reports"
Quick start
# 1) Pull the latest posts incrementally (cheap; only fetches since the last run)
node skills/follow-aleabito/scripts/analyze-mentions.js --incremental --include-replies --resume
# 2) See where her attention is moving right now
node skills/serenity-radar/scripts/radar.js --window 14 --top 12
Sample radar output (trimmed):
## 🔥 Heating (attention momentum — recent vs prior mentions)
ticker=JBL Δ=13 recent=25 prev=12
ticker=XFAB Δ=12 recent=12 prev=0 # new entrant + heating = emerging focus
ticker=SIVE Δ=10 recent=93 prev=83
## 🔄 Theme rotation
▲ Western supply chain / policy: +8 ▼ AI compute / neocloud: −12
Then ask your agent: "Analyze $SIVE like Serenity" → serenity-method produces the 5-block analysis. All output is 中文 by default, English on request.
Live dashboard
A self-contained, interactive lifetime research dashboard built on top of these skills — spanning her entire public archive, it maps attention momentum, mention structure (posts vs. replies vs. quotes), recent moves, and research priority across the tracked tickers, with linked price trends.
👉 Open the live dashboard — no login, no external dependencies. For tracking & research only, not investment advice.
How it was built
The skills are distilled from her complete public history, pulled with follow-aleabito's archive backfill (X API full-archive search). The empirical patterns in serenity-radar/references/patterns.md (theme-rotation logic, selection signature, catalyst playbook, conviction tells) are mined directly from those 6,120 posts — including the November-2025 drawdown (IREN −38% / NBIS −35%) she held through, which is why the radar's signals aren't overfit to a single up-only window.
Caveats & disclaimer
-
Candidate generator, not an oracle. The radar predicts her interest, not price or correctness.
-
Survivorship bias. Her archive over-weights names that worked; treat "she ramped X and it ran" as not evidence it repeats.
-
Single-account fragility. One person, one style, one era.
-
Not investment advice. For information tracking and research only. Do nothing with this that you wouldn't do after your own due diligence.
License
MIT.
中文
这是什么
三个环环相扣的技能,把一个高产的 X(推特)研究账号变成一套可复用的工作流:
技能 作用 关键命令
follow-aleabito
数据层。 用 X API 抓取她的发帖,生成小白友好的简报、累计提及分析,以及持久的研究地图;含全档案回填。
node skills/follow-aleabito/scripts/analyze-mentions.js --incremental --resume
serenity-method
她的分析方法,通用化。 把她的风格套到任意股票:关键卡点 / 供应链 OSINT 发现 → 第一性原理 → Buffett 五问质量门(默认 unverified)→ 叙事 vs 基本面卫生 → 研究地图 vs 可投资结论。
(按需调用:"用 Serenity 的方法分析 $X")
serenity-radar
她的注意力流向。 从提及档案算注意力动量(升温标的、新进、重仓核心、主题轮动),并按她的风格生成候选——候选发生器 + 检查清单,不是预言机。
node skills/serenity-radar/scripts/radar.js --window 14
管线: follow-aleabito(数据)→ serenity-method(她怎么推理)→ serenity-radar(她下一步可能看什么),后者建立在前者之上。
安装
- 把技能复制到你 agent 的技能目录:
cp -R skills/* ~/.codex/skills/ # Codex
# 或
cp -R skills/* ~/.claude/skills/ # Claude Code
-
脚本需要 Node 18+。
-
X API token —— 在
~/.follow-aleabito/.env里放一个 bearer token:
X_BEARER_TOKEN=你的token
全档案回填(--archive)需要你的 X API 项目开通全档案搜索(/2/tweets/search/all);日常增量抓取用标准时间线端点即可。
- 指定数据目录(分析 CSV 存放处):
export FOLLOW_ALEABITO_REPORTS_DIR="$HOME/aleabito-reports"
快速上手
# 1) 增量拉取最新发帖(很省;只抓上次之后的)
node skills/follow-aleabito/scripts/analyze-mentions.js --incremental --include-replies --resume
# 2) 看她现在注意力往哪走
node skills/serenity-radar/scripts/radar.js --window 14 --top 12
然后对你的 agent 说:"用 Serenity 的方法分析 $SIVE" → serenity-method 会产出五段式分析。所有输出默认中文,可按需出英文。
实时看板
一个基于这套技能制作、自包含的交互式 全周期(lifetime)研究看板——覆盖她的全部公开历史,把注意力动量、提及结构(发帖 vs. 回复 vs. 引用)、近期变化和研究优先级在所有跟踪标的上可视化,并联动价格趋势。
👉 打开实时看板 —— 无需登录、无外部依赖;仅作跟踪与研究用途,不构成投资建议。
数据怎么来的
这些技能蒸馏自她的全部公开历史,用 follow-aleabito 的全档案回填(X API 全档案搜索)抓取。serenity-radar/references/patterns.md 里的经验模式(主题轮动逻辑、选股签名、催化剂打法、重仓信号)直接从这 6,120 条帖子里挖出——其中包含她扛过的 2025 年 11 月回撤(IREN −38% / NBIS −35%),这也是雷达信号没有过拟合到"单边上涨窗口"的原因。
风险与免责
-
候选发生器,不是预言机。 雷达预测的是她的兴趣,不是价格或对错。
-
幸存者偏差。 她的档案天然偏向"成功的标的";别把"她加注 X 然后涨了"当成会重演的证据。
-
单账号脆弱性。 一个人、一种风格、一个时代。
-
不构成投资建议。 仅作信息跟踪与研究用途;请以你自己的尽调为准。
许可
MIT。
Built with the follow-aleabito · serenity-method · serenity-radar skills. Not affiliated with @aleabitoreddit. 与 @aleabitoreddit 无隶属关系。