llm-wiki-setup

$npx mdskill add daymade/claude-code-skills/llm-wiki-setup

帮用户搭一个**金融投研专用 LLM Wiki**(Karpathy 模式):纯 markdown 文件 + `[[wikilink]]` 互联 + LLM 维护,知识随用复利。

SKILL.md

.github/skills/llm-wiki-setupView on GitHub ↗
---
name: llm-wiki-setup
description: Co-create a personal investment-research LLM Wiki (Andrej Karpathy's pattern) where the user's OWN analysis framework becomes a living CLAUDE.md — by interviewing them, NOT by handing them a template. Use whenever the user wants to build a compounding research knowledge base, 投研第二大脑, 投研知识库, or 个人投研 wiki; instantiate Karpathy's LLM Wiki gist for finance/investing; turn their stock-picking, analyst-tracking, or earnings-watching workflow into a structured markdown vault; or build a wiki tracking companies / industries / macro / analysts over time. Pure markdown + wikilinks, NO RAG / vector DB (Karpathy's core idea — do not over-engineer). Also triggers for ingesting research reports / earnings calls / expert notes into an existing wiki, and for post-earnings prediction→fulfillment reviews. Core value = extracting the user's personal investment preferences into THEIR OWN schema, never imposing a standard one.
---

# LLM Wiki Setup(投研第二大脑共创)

帮用户搭一个**金融投研专用 LLM Wiki**(Karpathy 模式):纯 markdown 文件 + `[[wikilink]]` 互联 + LLM 维护,知识随用复利。

**但核心不是给一份投研模板——是引导用户把他自己的投资判断方式,提炼成他专属的 CLAUDE.md。**

## ★ 先读这一条(这个 skill 的灵魂)

**每个人用自己的语言、自己的投资偏好,建自己的 CLAUDE.md。**

两个投资者看同一家公司,关注点可能完全不同——一个看「下季度订单能否超市场预期」,另一个看「管理层电话会上的语气和信心」。**给他们同一份模板,就抹掉了让 wiki 有用的那个东西。**

- ✅ 你的工作 = 访谈用户 → 提炼他的关注维度 → 用**他的话**写进 CLAUDE.md
- ❌ 你的失败 = 套一份「标准投研 schema」让他填空,或让他照抄 `examples/`

`examples/investment-research-CLAUDE.md` 是**一个人长成的样子**,给用户看可能性,**禁止照抄**。它像模板一样被搬走,这个 skill 就失败了。

## 不碰的红线(Karpathy 原意,别 over-engineer)

纯 markdown + wikilink + grep。**不加 RAG / 向量库 / embedding。** 知识靠预编译进结构化页「复利」,不是每次 query 重新检索原始文档——这是本模式相对 RAG 的根本区别,也是 Karpathy 的核心 idea。别加回任何检索层,别加 knowledge graph / 自动 health-check 之类机制(社区有些版本加了,那是 over-engineer)。

## 机制层 vs 规则层(贯穿全程的区分)

| | 内容 | 处置 |
|---|---|---|
| **机制层** | 三层目录 + wikilink + lint + git hook | ✅ 通用工程结构,`scripts/init_vault.py` 直接装 |
| **规则层** | 看哪些维度 / 怎么记观点 / 要不要分析师归属 / 怎么复盘 / 要长报告还是三行 | ❌ 用户的投资大脑,**访谈长出来**,绝不给模板 |

机制层照抄没问题(它是 Karpathy 模式的工程卫生,跟「你怎么投资」无关)。规则层照抄 = 背叛方法论。

## 工作流

### Phase 0 — 判断意图
- **新建 vault** → Phase 1
- 已有 vault,**ingest 一份源** → 直接读 `references/ingest_sop.md`
- 已有 vault,**财报后复盘某标的** → `references/fulfillment_sop.md`
- **query** → 读 vault 的 `index.md` + 相关页,带 citation 综合答;好答案回填 synthesis

### Phase 1 — scaffold 机制层
```bash
python scripts/init_vault.py <目标目录>
```
建空骨架(三层目录 + lint + hook 占位 + 空 index/log + CLAUDE 骨架)。**这一步只装机制层,不写任何 schema。**

### Phase 2 — 访谈共创 CLAUDE.md ★核心步骤
**读 `references/interview.md`,按它的 8 个维度一条条访谈用户**,把回答用**他自己的话**写进 `<vault>/CLAUDE.md` 规则层的占位。

- 一次问一个维度,别一口气灌
- 用户不在乎的维度**直接砍**(极简 > 全面)
- 卡住才翻 `examples/` 给灵感,明说「别抄,挑你戳中的」
- 自检:写好的 CLAUDE.md 像不像「这个人」?**像通用模板就重来**

### Phase 3 — 启用防腐
```bash
cd <vault> && git init
git config core.hooksPath .githooks   # local 配置,换机/重 clone 要重设
PYTHONUTF8=1 uv run --no-project --with pyyaml python3 scripts/lint-vault.py wiki  # 确认绿灯
```

### Phase 4 — 首次 ingest 演示
拿用户**一份真实的源**(研报 / 电话会 / 纪要),按 `references/ingest_sop.md` 走一遍 HITL 5 卡点,让他亲眼看到 wiki 怎么从源长出来。**用用户自己的素材,不要用 examples。**

## 后续运营(按需读 references)

| 场景 | 读 |
|---|---|
| ingest 新源 | `references/ingest_sop.md`(doc_type 用用户自己定的分类) |
| 财报后复盘 | `references/fulfillment_sop.md`(分析师回测调 `analyst-track-record` skill,别重造) |
| vault 卫生(派生值漂移) | `references/prune_discipline.md` |
| 复盘页对抗审查 | `references/counter_review.md` |
| 怎么访谈提炼用户的投资大脑 | `references/interview.md`(Phase 2 的完整方法) |

## 为什么这个 skill 是 inline(不设 context: fork)

它要调 `analyst-track-record` skill(复盘回测)、跑 Bash(scaffold / lint)、可能并行 Task 取财报数据——subagent 不能调 skill 或 spawn subagent,所以必须 inline。

## Next Step

vault 搭好、用户开始 ingest 卖方研报后,如果他想回测某分析师过去准不准 → 建议接 `analyst-track-record` skill(双维度命中率,有 validated 脚本)。

More from daymade/claude-code-skills

SkillDescription
asr-transcribe-to-textTranscribes audio and video files to text using Qwen3-ASR. Supports two modes — local MLX inference on macOS Apple Silicon (no API key, 15-27x realtime) and remote API via vLLM/OpenAI-compatible endpoints. Auto-detects platform and recommends the best path. Triggers when the user wants to transcribe recordings, convert audio/video to text, do speech-to-text, or mentions ASR, Qwen ASR, 转录, 语音转文字, 录音转文字. Also triggers for meeting recordings, lectures, interviews, podcasts, screen recordings, or any audio/video file the user wants converted to text.
auto-repo-setup|
benchmark-due-diligence>
bigdata-skill>-
capture-screenProgrammatic screenshot capture on macOS. Find window IDs with Swift CGWindowListCopyWindowInfo, control application windows via AppleScript (zoom, scroll, select), and capture with screencapture. Use when automating screenshots, capturing application windows for documentation, or building multi-shot visual workflows.
claude-code-history-files-finderFinds and recovers content from Claude Code session history files. This skill should be used when searching for deleted files, tracking changes across sessions, analyzing conversation history, or recovering code from previous Claude interactions. Triggers include mentions of "session history", "recover deleted", "find in history", "previous conversation", or ".claude/projects".
claude-md-progressive-disclosurer|
claude-skills-troubleshootingDiagnose and resolve Claude Code plugin and skill issues. This skill should be used when plugins are installed but not showing in available skills list, skills are not activating as expected, or when troubleshooting enabledPlugins configuration in settings.json. Triggers include "plugin not working", "skill not showing", "installed but disabled", or "enabledPlugins" issues.
cli-demo-generatorGenerates professional animated CLI demos as GIFs using VHS terminal recordings. Handles tape file creation, self-bootstrapping demos with hidden setup, output noise filtering, post-processing speed-up, and frame-level verification. Use when users want to create terminal demos, record CLI workflows as GIFs, generate animated documentation, build demo tapes for README files, or need to showcase any command-line tool visually. Also triggers on "record terminal", "VHS tape", "demo GIF", "animate my CLI", or any request to visually demonstrate shell commands.
cloudflare-troubleshootingInvestigate and resolve Cloudflare configuration issues using API-driven evidence gathering. Use when troubleshooting ERR_TOO_MANY_REDIRECTS, SSL errors, DNS issues, or any Cloudflare-related problems. Focus on systematic investigation using Cloudflare API to examine actual configuration rather than making assumptions.