⚠️ Disclaimer: This tool is for learning and research purposes only. All generated content is AI style simulation,does not constitute any investment advice. Market data must be collected in real time,LLMs are prohibited from fabricating data such as price limits and capital flows。
📝 Overview
Distill 200+ articles from 5 WeChat public account stock investment influencers into structured style profiles (six-dimensional DNA: trading system / market judgment / expression style / content depth / interaction characteristics / hotspot map), and use an AI Agent to deliver multi-mode stylized analysis. Market data is collected in real time via WebSearch with source annotations for each value, and generated content is released only after quality audit (dual verification at the surface and thinking levels).
✨ Features
| Function Module | Capability Description | Core Value |
|---|---|---|
| Multi-Influencer Parallel Comparison | 5 influencers conduct stylized analysis of the same target/sector, outputting consensus and divergence | Multi-perspective cross-validation to avoid single bias |
| Daily Review | Generate a same-day market review article in a specified influencer's style | Understand market sentiment and trading logic from a stylized perspective |
| Sector Funnel | Four-layer screening under the influencer analysis framework: track → leader → certainty → action | Industry chain logic penetrates market hotspots |
| Short-Term Stock Selection | Extract individual stock mention frequency from popular WeChat public account articles, and screen by trading type | Quantified stock selection based on influencer attention |
| Portfolio Review | Retrieve influencers' historical views on each target, and evaluate holdings according to the trading framework | Stylized holding diagnosis and risk assessment |
| Financial Report Commentary | Interpret company financial reports in an influencer's style, with multi-dimensional fundamental analysis | Professional fundamental interpretation with stylized output |
🔑 Authentication
Go to redfox.hk to register and obtain an API Key, and configure it in the following way:
# 环境变量
export REDFOX_API_KEY="ak_你的密钥"
# 或配置文件 ~/.qoder/apis/redfox.json
{"api_key": "ak_你的密钥"}
API Key is used for:
- queryWorkList(article sync): charged per use, each influencer sync deducts 1 credit
- record/save(call record reporting): used only for statistics,no charge
Quick Start
Describe your needs directly in natural language, and the Agent automatically identifies the mode:
| Scenario | Trigger Example |
|---|---|
| Single-Influencer Analysis | "Analyze CATL in Mao Bi Dao's style" |
| Multi-Influencer Comparison | "Analyze Moutai separately in the styles of 5 influencers" |
| Sector Analysis | "Jointly analyze the aerospace sector" |
| Short-Term Stock Selection | "Screen semiconductor short-term picks in Cai Tang Ping's style" |
| Portfolio Review | "Review my holdings in Tou Zi Ming Jian's style: China Satellite, AECC Aviation Power" |
| Earnings Review | "Review Kweichow Moutai's latest earnings report" |
| Daily Recap | "Write today's recap in the style of Mao Bi Dao" |
| Viewpoint Tracking | "Check Cai Tangping's views on China Satellite" |
Preset Big V Profiles
profiles/ One JSON profile per Big V in the directory, ready to use out of the box:
| Big V | Trading Type | Core Style |
|---|---|---|
| Cai Tangping | Short-term + Trend | Macro trends + contrarian thinking, Daoist natural philosophy, "old fans all know" |
| Mao Bi Dao | Short-term | Multi-asset perspective + cycle reasoning, humorous and casual, "tonight's uncle sauce" |
| Gelan Investment Research | Long-term | Industry chain in-depth research report style, data-driven, "industry prosperity determines everything" |
| Investment Insight | Long-term | Pure technical trend following, index as the core, "trend is the best friend" |
| Lifelong Black & White | Long-term | Deep value investing, moat + margin of safety, "get rich slowly" |
Execution Process
Phase -1: Incremental Article Sync (ask when triggered)
On the first trigger of the daymust first ask the userwhether to sync the Big Vs' latest articles (each costs 1 credit).
Three choices: Skip / Partial sync (specify Big Vs) / Full sync (5 people, after completionmust enter Phase 5 supplementary distillation）。
python3 "$SKILL_PATH/scripts/analyze.py" --author "财躺平" --mode sync
Phase 0: Environment Preparation + AI Bias Self-Check
python3 "$SKILL_PATH/scripts/analyze.py" --check-env
After checking dependencies and profile readiness, assess information richness according to the following standards and write it at the beginning of the generated content:
| Level | Characteristics | Response Strategy |
|---|---|---|
| Level A(information abundant) | High-confidence profiles, complete market data | Reverse test: When Big Vs are bullish, why are smart people bearish? |
| Level B(information moderate) | Medium-confidence profiles, some data missing | Mark confidence level for each estimated data point |
| Level C(information scarce) | Low-confidence profiles, data missing | First-principles questioning, do not piece together a report that "looks complete" |
Bias self-check list:
- Does the feeling of "certainty" come from market data or the tone of the Big V profiles?
- If you remove the Big V profiles and look only at the market data, would the conclusion change?
- Have you only presented bullish reasons and ignored counterarguments?
Phase 1: Market Data Collection
Iron rule: market data must be collected in real time; LLM fabrication is prohibited.
Data acquisition: user manual input > WebSearch > LLM fabrication prohibited。
Cross-validate key data (≥2 sources, differences > 1% must be flagged):
| Data Item | Primary Source | Secondary Source |
|---|---|---|
| Number of limit-up/limit-down stocks | Exchange official website | East Money/Tonghuashun |
| Turnover | Exchange official website | Wind/THS |
| Northbound capital | Shanghai-Hong Kong-Shenzhen Stock Connect official website | East Money |
| Main capital flow | East Money | THS |
Phase 2: Style profile loading
python3 "$SKILL_PATH/scripts/analyze.py" --list-profiles
Phase 3: Content generation
Data source disclosure (mandatory for all scenarios): After each market data point, indicate the source domain in parentheses (e.g., "THS 10jqka.com.cn"); writing only "AI inference" is prohibited; add a correction prompt at the end of the article: "If the data is inaccurate, please provide the real data, and I will regenerate it."
Quality audit rules: Surface features (signature expressions/trading system/article structure/market indicators/tone/terminology) weight 50%, thinking-layer features (core thinking/argumentation style/philosophical values) weight 50%. A composite score > 70 passes; < 70 is rejected and regenerated.
Mode command quick reference：
| Mode | CLI command | Output |
|---|---|---|
| daily | --author "财躺平" --mode daily |
output/{大V}_复盘_{日期}.md |
| team | --mode team --authors "财躺平,格兰投研" |
output/多V对比_{日期}.md |
| sector | --author "财躺平" --mode sector --sector "航天" |
output/{大V}_{板块}分析_{日期}.md |
| track | --author "财躺平" --mode track --stock "中国卫星" |
output/{大V}_{标的}观点跟踪_{日期}.md |
| screen | --author "财躺平" --mode screen --sector "半导体" |
output/{大V}_选股_{日期}.md |
| portfolio | --author "财躺平" --mode portfolio --stocks "A,B,C" |
output/{大V}_组合复盘_{日期}.md |
| earnings | --author "格兰投研" --mode earnings --stock "贵州茅台" |
output/{大V}_{标的}财报_{日期}.md |
| sync | --author "财躺平" --mode sync |
output/{大V}_增量同步_{日期}.md |
Phase 4: Quality audit
python3 "$SKILL_PATH/scripts/quality_audit.py" --article "output/财躺平_复盘_2026-01-15.md" --author "财躺平"
Hard compliance rules (any violation is directly rejected)：
| # | Rule | Prohibited behavior |
|---|---|---|
| 1 | Do not promise returns | "guaranteed to make money", "guaranteed profit without loss" |
| 2 | Do not predict specific return rates | Promise specific numbers |
| 3 | Do not imply insider information | Imply having insider information |
| 4 | Mark data sources | Mark "based on public data" |
| 5 | Mark AI simulation | "Unrelated to the original big V, only an AI style simulation" |
| 6 | No extreme instructions | "Buy xxx with all your capital" |
All content must be marked "AI style simulation, not investment advice"; do not impersonate big Vs; do not directly recommend stocks.
Phase 5: Incremental profile update
Iron rule: after each sync, supplementary distillation must be completed; it must not be skipped.
Trigger condition: in Phase -1, the user chose sync.
Execution process:
- Read the incremental sync report
- Compare item by item the content that needs updating (new signature expressions/new stocks/new indicators/analysis framework changes/word count changes) → modify the JSON, update the distillation date and number of sample articles
- If there are no updates, clearly state the reason
- Mark significant changesStyle drift warning
- Perform a full verification of 5 profiles once a month
Other resources
- Quality standards details:Quality pass standards.md
- Article template:article_template.md
- Big V profiles:
profiles/{大V名}_profile.json
Dependencies
pip install requests