AI-Trader Market Intel
HKUDS/AI-Trader
Reads AI-Trader's read-only market snapshots, grouped financial news and events board through its market-intel endpoints, for context before trading or posting.
Plans and writes a three-to-eight-part long-form article series that breaks down one company, built on fact-checked financials, valuation and management analysis.
SKILL.md written in Chinese; this summary is our English description.
Install Deep Company Article Series
$ npx skills add xbtlin/ai-berkshire --skill deep-company-series -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install xbtlin/ai-berkshire deep-company-series --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/xbtlin/ai-berkshire.git skills-src && mkdir -p .claude/skills && cp -r skills-src/codex-skills/deep-company-series .claude/skills/deep-company-series && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "deep-company-series" agent skill from https://github.com/xbtlin/ai-berkshire/tree/main/codex-skills/deep-company-series into .claude/skills/deep-company-series/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "deep-company-series", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/xbtlin/ai-berkshire/tree/main/codex-skills/deep-company-seriesType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add xbtlin/ai-berkshire --skill deep-company-series -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install xbtlin/ai-berkshire deep-company-series --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/xbtlin/ai-berkshire.git skills-src && mkdir -p .agents/skills && cp -r skills-src/codex-skills/deep-company-series .agents/skills/deep-company-series && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "deep-company-series" agent skill from https://github.com/xbtlin/ai-berkshire/tree/main/codex-skills/deep-company-series into .agents/skills/deep-company-series/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "deep-company-series", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add xbtlin/ai-berkshire --skill deep-company-series -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install xbtlin/ai-berkshire deep-company-series --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/xbtlin/ai-berkshire.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/codex-skills/deep-company-series .cursor/skills/deep-company-series && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "deep-company-series" agent skill from https://github.com/xbtlin/ai-berkshire/tree/main/codex-skills/deep-company-series into .cursor/skills/deep-company-series/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "deep-company-series", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/xbtlin/ai-berkshire.git --path codex-skills/deep-company-series--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add xbtlin/ai-berkshire --skill deep-company-series -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install xbtlin/ai-berkshire deep-company-series --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/xbtlin/ai-berkshire.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/codex-skills/deep-company-series .gemini/skills/deep-company-series && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "deep-company-series" agent skill from https://github.com/xbtlin/ai-berkshire/tree/main/codex-skills/deep-company-series into .gemini/skills/deep-company-series/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "deep-company-series", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install xbtlin/ai-berkshire deep-company-seriesInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add xbtlin/ai-berkshire --skill deep-company-series -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/xbtlin/ai-berkshire.git skills-src && mkdir -p .github/skills && cp -r skills-src/codex-skills/deep-company-series .github/skills/deep-company-series && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "deep-company-series" agent skill from https://github.com/xbtlin/ai-berkshire/tree/main/codex-skills/deep-company-series into .github/skills/deep-company-series/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "deep-company-series", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add xbtlin/ai-berkshire --skill deep-company-series -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install xbtlin/ai-berkshire deep-company-series --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/xbtlin/ai-berkshire.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/codex-skills/deep-company-series .opencode/skills/deep-company-series && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "deep-company-series" agent skill from https://github.com/xbtlin/ai-berkshire/tree/main/codex-skills/deep-company-series into .opencode/skills/deep-company-series/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "deep-company-series", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
Facts
- Skill name
deep-company-series- GitHub stars
- 17k
- Token cost
- ~2k tokens
- SKILL.md length
- 727 words
- Files
- 1
- Skills in repo
- 22
- Repo updated
- First seen
- Licence
- MIT
At a glance
Plans and writes a three-to-eight-part long-form article series that breaks down one company, built on fact-checked financials, valuation and management analysis.
- Works in 4 steps: 先核查事实（不要直接改） → 判断修订级别 → 修订后联动检查 → …
- Writing a multi-part public deep dive on a single company
- SKILL.md covers Codex adapter note, 一、触发场景, 二、系列篇目模板 and 二·五、篇内结构模板（从《看懂腾讯》8 篇正文提炼）, plus 7 more sections
- Calls git and python3
What it does
The target is a public series, for channels such as WeChat, aimed at readers who want a textbook-level understanding of a company, with each article readable alone but sharing one view on valuation, management and price. The number of parts depends on complexity: seven or eight for a multi-business company with a large investment portfolio, four to six for a middle case, and three for a company with a clear core business, where the eight standard themes are folded into three articles but still all covered.
The skill sets research discipline: confirm today's date with the `date` command and state the data cutoff in the report header, cross-check financial data, use exact arithmetic tools such as `tools/financial_rigor.py` for valuation, and label uncertainty and source gaps. It is not for single research reports, earnings commentary or industry studies, which have their own commands. The text is in Chinese.
When your agent uses it
- Writing a multi-part public deep dive on a single company
- Deciding how many articles a company's complexity justifies
- Fact-checking and cross-validating financial claims before publishing
Example prompts
- “Plan a three-part article series on a streaming company, covering the moat, hidden assets and valuation.”
- “Decide how many parts a conglomerate with several business lines needs, and outline each one.”
- “Write the valuation and decision article for the series with exact arithmetic for every figure.”
Requirements
- The ai-berkshire repository's `tools/` folder, such as `financial_rigor.py`
- Python 3
- Web search, for current financial data
Workflow steps
4 steps, taken from the step headings in SKILL.md.
What it can do on your machine
Read from SKILL.md and the folder at commit a221a20. It shows what the files ask for, not the result of running them.
Tool permissions
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Runs code
Shell commands in SKILL.md call:
gitpython3
From the folder's file list and the shell code blocks in SKILL.md.
Network
No URLs in SKILL.md. Its commands use git, which can reach the network depending on how they are called.
From URLs in SKILL.md, links to its own repository left out.
Credentials
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Context cost
Deep Company Article Series loads about 2k tokens when it runs. Until then it costs about 27 tokens; SKILL.md has 727 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
Safety
Auto-check passed
The automated check found no risky patterns in SKILL.md.
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
SKILL.md
The full file from xbtlin/ai-berkshire at commit a221a20, republished under its MIT licence (© xbtlin). 727 words, ~2,018 tokens.
Download SKILL.mdSave it as
.claude/skills/deep-company-series/SKILL.md (or your agent's skills folder).- name
- deep-company-series
- description
- AI Berkshire skill: 看懂XX公司（深度公司系列）：3-8 篇长文拆一家公司. Source: skills/deep-company-series.md.
Codex adapter note
This skill is generated from skills/deep-company-series.md so Claude Code and Codex users share one canonical workflow.
- Treat
$ARGUMENTSas the user's request in the current Codex thread. - When the source mentions Claude-only surfaces such as Task, Agent, WebSearch, Bash, Read, or Write, use the closest Codex capability available in this session: subagents when available, web search when needed, shell commands for local tools, and normal file edits for workspace files.
- Use shared project tools from
tools/in this repository. Prefer running commands from the repository root with paths likepython3 tools/financial_rigor.py ...; if the current thread starts outside the repo, locate the actual checkout path first instead of assuming a fixed home-directory path. - Before starting research, run the
datecommand to confirm today's date; treat it as the baseline for "latest" data and state the data cutoff date in the report header. Never assume the current date from training data. - Preserve the research quality rules from
AGENTS.md: cross-check financial data, use exact arithmetic tools for valuation/math, and clearly label uncertainty and source gaps.
看懂XX公司（深度公司系列）：3-8 篇长文拆一家公司
为 $ARGUMENTS 撰写一个《看懂XX》深度长文系列（3-8 篇，按公司复杂度定，见"篇数适配"），发布在公众号/视频号等公开渠道。核心 IP 不是"会写"，而是"会改"——99% 的财经文章在违反本 skill 的事实核查标准。
参考样本：reports/腾讯/《看懂腾讯》/
一、触发场景
用户希望为一家公司做"教科书级别"的深度研究，并以系列长文形式公开发布。区别于一篇研报：
- 3-8 篇（复杂公司 8 篇约 12 万字，简单公司 3 篇约 3 万字），从认知重置到决策框架完整闭环
- 每篇独立成文（适合单篇分享），但贯穿一套估值/管理层/价格判断
- 写给"愿意花 90 分钟读懂一家公司"的读者，不是写给券商客户
不适合用本 skill 的场景：单篇研报、季报点评、行业研究——那些用 /investment-research、/earnings-review、/industry-research。
二、系列篇目模板
篇数适配公司复杂度（先定篇数，再定篇目）
篇数不固定为 8。 8 篇是腾讯这种"多业务 + 万亿投资组合 + 20 年管理层故事"的公司才撑得起的容量。写之前先问：这家公司有几个能用"一个尖锐问题"独立成文的主轴？有几个写几篇。
| 复杂度 | 特征 | 篇数 | 例子 |
|---|---|---|---|
| 高 | 多条业务线各自成生意 + 隐藏资产/投资组合 + 管理层史料丰富 | 7-8 篇 | 腾讯 |
| 中 | 2-3 条业务线 + 一个重大时代变量 | 4-6 篇 | — |
| 低 | 主业清晰、核心问题少而集中 | 3 篇 | 快手 |
3 篇的标准合并方式（下面 8 主轴仍是检查清单，全部要覆盖，只是折进 3 篇）：
| # | 篇名方向 | 折入的主轴 |
|---|---|---|
| 01 | 你以为你看懂了 X（开篇 + 生意与护城河） | 认知重置 + 护城河 + 利润引擎概览 |
| 02 | X 的最大变量（当下叙事/隐藏资产） | 隐藏资产 + AI/时代变量 |
| 03 | 多少钱值得买，什么信号必须卖（财务 + 决策终章） | 财务深度 + 管理层要点 + 估值决策 |
判断信号：如果某篇写到一半发现在"凑内容"（同一事实换说法重复出现），说明篇数定多了——合并。反之，某篇超过 12,000 字还塞不下，拆开。
8 主轴全模板（复杂公司用；简单公司当检查清单折叠）
| # | 篇名模板 | 核心问题 | 字数 |
|---|---|---|---|
| 01 | 你以为你看懂了 X，其实没有 | 认知重置：破 3 个常见错觉 | 4,000-5,000 |
| 02 | X 的护城河——<生意本质一句话> | 护城河深不深、未来 5/10 年还在不在 | 6,000-8,000 |
| 03 | X 的最大利润引擎——<最赚钱业务> | 主业是什么、为什么能持续 | 6,000-8,000 |
| 04 | X 藏在账上的另一家公司——<隐藏资产> | 投资组合 / 子公司 / 隐藏价值 | 8,000-10,000 |
| 05 | AI（或当下叙事）时代，X 是赢家还是输家 | 时代变量：分业务拆 AI 影响 | 8,000-10,000 |
| 06 | 用巴菲特方式拆 X 的财报 | 财务深度：毛利率/FCF/ROE/SBC | 8,000-10,000 |
| 07 | <管理层金句>——X 的管理层值不值得托付 | 资本配置纪律 + 诚信检验 + 接班人 | 8,000-10,000 |
| 08 | 多少钱值得买，什么信号必须卖（系列终章） | DCF 三情景 + 红线清单 + 仓位框架 | 10,000-12,000 |
加一篇 00-系列说明.md 作为目录索引，不发表。
篇目必须适配公司：模板是腾讯长出来的，不是套子。某篇没有独立内容（如没有投资组合的公司写不了 04），就替换成该公司特有的主轴（如快手的可灵 AI、电商），或按上面的合并方式折进更少的篇数。判断标准：这一篇能否用"一个尖锐问题"独立成文。
二·五、篇内结构模板（从《看懂腾讯》8 篇正文提炼）
每篇不是自由发挥，有固定骨架。共性结构：
- 头部引用块：
> 《看懂X》系列 · 第 0N 篇 · <主轴>，开篇加阅读时间 - 第一节永远是钩子：一组反常识数字 / 一条反常曲线 / 一个让市场担心的数字，不写背景铺垫
- 每篇末尾：
## 本篇要点回顾（5-8 条，数字与正文严格一致）+## 下期预告（用尖锐问题预告，不剧透结论） - 结尾两行斜体：
*本文是《看懂X》系列第 0N 篇...*+*本系列基于公开信息和价值投资框架研究，不构成任何投资建议。*
各篇专属骨架：
| 篇 | 骨架 |
|---|---|
| 01 开篇 | 股价被市场反复"重新定性"的历史曲线 → 三个常见错觉（每个用财报数据破）→ "需要同时戴几副眼镜"阅读地图表（链接到 02-08）→ 估值钩子数字表（只留悬念不给结论） |
| 02 护城河 | 挑战者失败名单（有名有姓有年份）→ 迁移成本的本质（用户迁不走什么）→ 生态飞轮 → 5 年/10 年后护城河还在吗 → 一个反向思维实验（"如果我有 1000 亿能不能复制它"） |
| 03 利润引擎 | 反常识数字开场 → 核心产品为何长青（拆到定价权层面）→ 增长空间量化（份额路径 + 反面论据：故事不工作的可能）→ 组织能力（为什么能持续出产品）→ 市场为什么低估 → AI/时代变量对本业务的影响 → 5 年后图景 |
| 04 隐藏资产 | TOP 持仓表（只统计未并表部分，并表的单独说明防双算）→ 会计处理差异（哪些进利润表哪些不进）→ "不会卖 vs 可能减持"分类 → 市场为何打折价（逐条）→ 几个意外发现 → 估算折扣的简单方法 |
| 05 AI/时代变量 | 让市场担心的数字 → 分业务拆赢家/输家（严重分化，不给整体结论）→ 关键反问（"X 输了单点之争，还能赢吗"）→ 5-10 年被绕过的场景推演（逐条反驳或承认）→ 5 个情景 → 未来一年监控信号清单 |
| 06 财务 | 一条反常曲线开场 → 利润表（毛利率变化归因）→ 口径选择（Non-IFRS vs GAAP，为什么）→ EPS 与回购 → 资产负债表（净现金）→ 现金流与 FCF Yield → 股东回报按"1 美元测试"→ ROE 变化是好是坏 → 估值横向+历史比较 → 价值投资 10 项 Checklist 逐条过 → 把财报压缩成三句话 |
| 07 管理层 | 核心团队表 → 关键人物的长期贡献（用具体决策，不用形容词）→ 利益对齐（持股/薪酬占利润比）→ 诚信检验：挑一次真实危机看应对 → 5 年承诺兑现率（逐条核对当年原话）→ 资本配置纪律 → 接班人隐忧（不回避）→ "买人三问"（诚实/能干/热爱）→ 综合评分 |
| 08 决策 | SOTP 分部估值（说清"你买的到底是什么"）→ 历史估值分位 + 必须警告"后视镜不是导航仪" → 未来利润三情景表（乐观/中性/保守 + 隐含增速，标明"是预测不是事实"）→ 红线清单（触发≠必卖，但必须重审）→ 赔率思维收尾 |
Show full SKILL.md (276 more words)Show less
三、写作风格规范
语气
- 直接、犀利、不说废话——第一句就给数字或反常识结论
- 价值投资框架——巴菲特/芒格/段永平/李录视角穿插（但不堆砌名言）
- 不预设立场——先摆数据、再推逻辑、最后得结论
- 呈现正反两面——每个核心判断都附"但另一方面..."的反方
- 公众号体感——前 18-20 字必须能独立站住（手机预览）
禁用词
| 禁用 | 原因 | 替代 |
|---|---|---|
| 显然 / 必然 / 一定 | 主观绝对化 | 数据显示 / 证据表明 |
| 我认为 / 我觉得 | 主观腔调 | 删除或改为"按本框架" |
| 教科书级别 / 神来之笔 | 流量党褒奖 | 描述具体事实 |
| 严重不匹配 / 严重低估 | 强主观词 | 给具体折让百分比 |
| 完美 / 无可挑剔 | 单边判断 | 加上反方观察 |
标题风格
- 用反差数字或反共识结论做钩子（"15 年 7 次挑战全失败"、"年薪 4292 万仅占利润的万分之 1.65"）
- 副标题中性、概括内容（"——
<本质判断>"） - 避免流量党比喻："小巴菲特"、"中国版 X"、"YYDS" 一律避开
- 用专业读者熟悉的术语（"伯克希尔"而不是"巴菲特"，公司名优于人名）
四、严苛事实核查 Checklist（核心 IP）
写之前就要警惕的"伪精确"陷阱
- 概率加权期望值：
30% × A + 50% × B + 20% × C = 期望 +X%这种计算几乎全是垃圾——概率分配是纯主观，给读者错误精确感。只列情景 + 触发条件 + 方向，不算加权期望。 - 第三方测算 MAU/份额：QuestMobile/七麦/CBNData 等口径差异巨大（同一时点能差 2-3 倍）。只用最可信的两个对比作 anchor，其他做定性描述。
- 历史增速线性外推：
2025 年 +33% × 5 年复合 → 2030 年 X是金融文盲式预测。情景假设 + 高/低区间 + 不是承诺。 - 未公开的持股比例：字节、Halti 类未上市公司持股从未公开披露。给区间，标"不可知"。
- 强归因：竞争对手失败 = 因为 X。多重原因都列出来，本文不做单一归因。
修订时必跑的 7 项检查
□ 1. 跨篇数字一致性：总市值、Non-IFRS 净利润、关键持股 % 全系列对齐
□ 2. 口径标注：Non-IFRS / GAAP / Non-IFRS-SBC / FCF 各用哪个，全文清楚
□ 3. 重复加计扫描：已并表子公司不在"投资组合"里、SOTP 不双算
□ 4. 横向比较公平性：不能"主业 PE（剔除现金+组合）" vs "对手 PE（不剔）"
□ 5. 概率加权全删：见上一条
□ 6. 绝对化表述全弱化：grep "显然|必然|严重|教科书|完美"
□ 7. 第三方数据来源标注：每条非财报数据后跟"（来源：X）"模型偏好
写之前先列出已知硬错误风险：
- 历史回报倍数：必须用累计投入口径（如 Riot 33 倍 不是 58 倍）
- 持股比例：必须看最新富途/财报口径（如腾讯持有美团 1.5% 不是 6.4%）
- "派息分派"会计处理：视同处置收益按 IFRIC 17 在宣派日确认（如京东在 2021，美团在 2022 但金额小）
- 总股本会反弹：SBC 集中年初授予会让股本短期上升
五、执行流程
阶段 1：调研（写 01-02 篇前完成）
- 阅读公司近 5 年年报、最新季报
- 阅读至少 3 份独立卖方研报（找共识 + 反共识）
- 用
/investment-team或/investment-research先生成内部研究底稿 - 与用户确认 8 篇的核心论点（避免写完才发现方向不对）
阶段 2：写作（按 01→末篇顺序写，不跳）
- 每篇写完先存
reports/{公司名}/《看懂{公司名}》/0X-XX.md - 目录冲突规则：若
reports/{公司名}/《看懂{公司名}》/已存在（旧版系列），不覆盖、不混放——新系列写入带日期后缀的新文件夹reports/{公司名}/《看懂{公司名}》-{YYYYMMDD}/，旧文件夹原样保留 - 不立即推 GitHub——等用户审阅
- 用户提修订意见后修改
- 修订完才 git push
阶段 3：跨篇一致性扫描（全部篇目写完后）
派 Explore agent 并行扫描全系列做以下检查：
- 同一数字（市值、净利润、持股比例）跨篇是否一致
- 同一术语（FBS、SBC、Non-IFRS）首次出现是否解释
- 引用关系：02 篇说"详见 06 篇"是否真的对应
- 要点回顾 vs 正文是否数字一致
阶段 4：发布前最终核查
# 推送前必须本地 grep 一次（按 ai-berkshire 隐私规则）
grep -r "<本机用户名>\|/Users/\|<个人身份信息>" reports/ | head确认无误后才 git pull --rebase && git commit && git push。
六、修订意见处理流程
用户给修订意见时，按以下顺序处理：
1. 先核查事实（不要直接改）
如果用户说"X 数据不对"，先用 Bash/Read 找原始数据交叉验证：
- 看 ai-berkshire 项目里同公司的 earnings/财报报告
- 看富途/官方披露
- 给出"用户说的数据 vs 我查到的数据 vs 我之前用的数据"三方对比
2. 判断修订级别
| 级别 | 类型 | 处理 |
|---|---|---|
| 🔥 硬错误 | 数字错、归因错、口径错 | 必改，不需犹豫 |
| ⚠️ 主观化 | 强主观词、绝对化、流量党比喻 | 弱化或删除 |
| 🔬 颗粒度 | 来源标注、口径细化 | 优先级低，按可读性平衡 |
| ❓ 不可靠 | 第三方测算差异大 | 删比改更稳（用户明确指示） |
3. 修订后联动检查
修一处先想"哪些地方还会引用这个数字/概念"。例：
- 改了总市值 → 全系列联动改 PE / 主业 PE / 折让 / FCF Yield
- 改了持股 % → 改 TOP 10 排序 + 历史持股表 + 减持清单
- 改了术语口径 → 改首次定义 + 后续引用 + 要点回顾
4. 推送后立即报告
推送成功（commit hash）。
[N] 处修订总结 [带表]：
- 改了什么
- 联动改了什么
- 还有什么没改
下一步等指示。七、本 skill 不做什么
- 不替读者做投资决策——所有篇章末尾"不构成投资建议"
- 不预测股价——只给"情景 + 触发条件"
- 不算"期望年化回报"加权值——主观概率分配会误导读者
- 不写"X 大佬也持有" —— 用别人的持仓为自己的判断背书是反价值投资的
- 不强求 8 篇都写——如果某篇没足够独立内容（如某公司管理层不够特别），合并到其他篇或减篇数
八、合规与隐私
- 所有公开报告只用公开信息（财报、官方披露、券商研报、知名第三方机构）
- 不用任何用户个人信息（公司花名、内部 IM、未公开持仓信息）
- 推送前必须用 grep 扫描 本机用户名 /
/Users// 真实姓名 等隐私字段 - 公开署名按用户多层身份策略，不混用
一句话总结
写《看懂 X 系列》的核心能力 ≠ 写得好，而是改得严—— 89% 的财经长文死于伪精确数字、主观加权期望值、绝对化表述。本 skill 的存在就是为了把这些坑全部标记出来，写之前避开，写之后扫干净。
© xbtlin, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
Files
Just SKILL.md in codex-skills/deep-company-series of xbtlin/ai-berkshire.
Open the folder on GitHubat commit a221a20
Compare with similar skills
Deep Company Article Series next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Deep Company Article Series this skillxbtlin/ai-berkshire | 17k | — | ~2k | Automated safety check: Pass | MIT | |
| AI-Trader Market IntelHKUDS/AI-Trader | 23k | — | ~1.1k | Automated safety check: Pass | None | |
| Eastmoney Market DataHKUDS/Vibe-Trading | 35k | — | ~1k | Automated safety check: Pass | MIT | |
| Stock Deep Analysis Workflowwbh604/UZI-Skill | 7.1k | — | ~9.1k | Automated safety check: Notes | MIT | |
| Zhengxi Fund Manager Views Librarylyra81604/zhengxi-views | 1.8k | — | ~1.6k | Automated safety check: Pass | MIT | |
| SEC EDGAR Filings FetcherHKUDS/Vibe-Trading | 35k | — | ~1.4k | Automated safety check: Pass | MIT |
Similar skills
Eastmoney Market Data
HKUDS/Vibe-Trading
Index of Eastmoney's free, no-token market data interfaces for China A-shares and Hong Kong stocks: fund flows, dragon-tiger lists, margin trading, reports and news.
Stock Deep Analysis Workflow
wbh604/UZI-Skill
Runs a staged deep analysis of a single stock on China A-share, Hong Kong and US markets, ending in an HTML report with valuation models and investor-panel scores.
Zhengxi Fund Manager Views Library
lyra81604/zhengxi-views
Answers questions with sourced quotes from one Chinese fund manager's public writings, applies his stated investment method and compares his words with real fund holdings.
SEC EDGAR Filings Fetcher
HKUDS/Vibe-Trading
Fetches U.S. SEC EDGAR data: resolves tickers to CIK numbers, lists recent 10-K, 10-Q and 8-K filings with document URLs, and pulls XBRL financial series.
Longbridge Earnings
helsome/folio
Earnings analysis — pre- and post-earnings. An agent skill from helsome/folio.
More from xbtlin/ai-berkshire
All 22 skills in this repoSupply Chain Bottleneck Hunter
xbtlin/ai-berkshire
Scans a long-running industry trend for supply chain chokepoints, aiming to find second- and third-layer suppliers that the market has not yet priced in.
Earnings Report Deep Reading
xbtlin/ai-berkshire
Reads a company's filings and earnings call material in depth, rates how complete the sources are and extracts the key figures into a structured review.
Earnings Report Deep-Read Team
xbtlin/ai-berkshire
Runs four parallel analyst personas over one earnings report, then an editor and reader-review pass turn the findings into a publishable article.
Era Alpha Investment Research Framework
xbtlin/ai-berkshire
A four-step research framework for finding and tracking high-growth core companies in one industry: map the sector, ask core questions, verify, then hold to the turning point.
Financial Data Cross-Validation
xbtlin/ai-berkshire
A research rule set for pulling company financials from prioritized sources by market and cross-checking every key figure against two independent sources.
Income Investment Analysis
xbtlin/ai-berkshire
Assesses whether a company's distributions are durable enough to earn a place in an income portfolio, starting from a ticker or company name.
Related
Categories
Questions about Deep Company Article Series
What does Deep Company Article Series do?
Plans and writes a three-to-eight-part long-form article series that breaks down one company, built on fact-checked financials, valuation and management analysis. The target is a public series, for channels such as WeChat, aimed at readers who want a textbook-level understanding of a company, with each article readable alone but sharing one view on valuation, management and price. The number of parts depends on complexity: seven or eight for a multi-business company with a large investment portfolio, four to six for a middle case, and three for a company with a clear core business, where the eight standard themes are folded into three articles but still all covered.
When should I use Deep Company Article Series?
Deep Company Article Series fits situations like: writing a multi-part public deep dive on a single company; deciding how many articles a company's complexity justifies; fact-checking and cross-validating financial claims before publishing.
How do I install Deep Company Article Series in Claude Code?
Run `npx skills add xbtlin/ai-berkshire --skill deep-company-series -a claude-code`. Or copy the skill folder (codex-skills/deep-company-series in xbtlin/ai-berkshire) into .claude/skills/deep-company-series in your project. Claude Code loads it when a task matches its description.
How do I install Deep Company Article Series in Codex?
Run `npx skills add xbtlin/ai-berkshire --skill deep-company-series -a codex`. Or copy the skill folder (codex-skills/deep-company-series in xbtlin/ai-berkshire) into .agents/skills/deep-company-series in your project. Codex loads it when a task matches its description.
Can I use Deep Company Article Series in Cursor, Gemini CLI or GitHub Copilot?
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add xbtlin/ai-berkshire --skill deep-company-series -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/deep-company-series, .gemini/skills/deep-company-series, .github/skills/deep-company-series and .opencode/skills/deep-company-series in your project.
What does Deep Company Article Series need to run?
Going by SKILL.md and its folder, Deep Company Article Series needs the command-line tools its instructions call (git and python3). Our summary lists: The ai-berkshire repository's `tools/` folder, such as `financial_rigor.py`; Python 3; Web search, for current financial data.
Does Deep Company Article Series access the network?
SKILL.md contains no URLs. Its commands use git, which can reach the network depending on how they are called. This is read from the text; nothing was executed.
Is Deep Company Article Series safe to install?
Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.
What licence does Deep Company Article Series use?
Deep Company Article Series is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.
How many tokens does Deep Company Article Series use?
About 2k tokens (SKILL.md is roughly 8.1k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.
What are the alternatives to Deep Company Article Series?
Skills that share tags, products or a category with Deep Company Article Series: AI-Trader Market Intel (HKUDS/AI-Trader, 23k stars), Eastmoney Market Data (HKUDS/Vibe-Trading, 35k stars), Stock Deep Analysis Workflow (wbh604/UZI-Skill, 7.1k stars) and Zhengxi Fund Manager Views Library (lyra81604/zhengxi-views, 1.8k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
Who maintains Deep Company Article Series?
xbtlin (a GitHub user) maintains it in xbtlin/ai-berkshire, which has 16,676 GitHub stars. The repository holds 22 skills in this directory. The repository was last updated on October 8, 2026.
Source: xbtlin/ai-berkshire on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.