alpha-zoo
Skill3 starsChanged 35 days agoMITView on GitHub
When the user asks about prebuilt cross-sectional alphas — Kakushadze 101, GTJA 191, Qlib 158, Fama-French / Carhart — or wants to bench a whole zoo on an investable universe (CSI 300, S&P 500, BTC-USDT, ...), this skill orients you. The zoo is the curated library; the bench is the evaluator. Counts are nominal; check alpha_zoo action=health for the live count currently loaded.
In this repository33 other files it gives its agents.
- llms.txtllms.txt
- breadth-chart-analystSkill
- canslim-screenerSkill
- factor-mining-pandaaiSkill
- backtesting-bias-avoidanceSkill
- backtest-reviewSkill
- data-quality-auditSkill
- paper-reproSkill
- pine-backtesterSkill
- stats-test-pickerSkill
- vectorbt-backtestSkill
- vectorbt-optimizeSkill
- akquantSkill
- ashare-pre-st-filterSkill
- cninfo-disclosureSkill
- ctp-apiSkill
- ptrade-strategy-writingSkill
- quant-buddySkill
- ssquant-futuresSkill
- tdxquantSkill
- tqsdk-tradingSkill
- wencai-screenerSkill
- defillama-dataSkill
- defillama-flowsSkill
- defillama-yieldsSkill
- perp-funding-basisSkill
- pine-debuggerSkill
- pine-developerSkill
- edgar-financialsSkill
- llmquant-equitiesSkill
- llmquant-macroSkill
- llmquant-riskSkill
- openquant-tradingSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 16
When the user asks about prebuilt cross-sectional alphas — Kakushadze 101, GTJA 191, Qlib 158, Fama-French / Carhart — or wants to bench a …
- Very specific: line 49
Universe loaders may not be wired for every market yet: When alpha_bench returns universe loader for X not yet implemented, that's the W2 …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
<!--
Imported from: https://github.com/HKUDS/Vibe-Trading
Source Path: agent/src/skills/alpha-zoo/SKILL.md
Curated for awesome-quant-skills
-->name: alpha-zoo description: Browse and bench the bundled alpha zoos — prebuilt cross-sectional factor libraries (Kakushadze 101, GTJA 191, Qlib 158, Fama-French / Carhart). Use when the user asks "which alphas exist", wants metadata on a named alpha, or wants to run IC/IR on a whole zoo over a universe. category: research
Alpha Zoo
Purpose
When the user asks about prebuilt cross-sectional alphas — Kakushadze 101, GTJA 191, Qlib 158, Fama-French / Carhart — or wants to bench a whole zoo on an investable universe (CSI 300, S&P 500, BTC-USDT, ...), this skill orients you. The zoo is the curated library; the bench is the evaluator.
Tools Available
| Tool | When to use |
alpha_zoo | Browse the library. action=list_alphas to enumerate (filterable by zoo / theme / universe), action=get_alpha for one alpha's metadata, action=health for registry load status. |
alpha_bench | Run IC / IR on one alpha or a whole zoo over a universe + period. Emits an HTML report. |
factor_analysis | Ad-hoc factor evaluation from a user-supplied factor CSV + return CSV. Use this when the user has their own factor (not in the zoo). |
Decision Tree
- "list all momentum alphas" →
alpha_zoowithaction=list_alphas, theme=momentum. - "show me gtja191_alpha_001" →
alpha_zoowithaction=get_alpha, alpha_id=gtja191_alpha_001. - "bench all of GTJA 191 on CSI 300 from 2020 to 2024" →
alpha_benchwithzoo=gtja191, universe=csi300, period=2020-2024. - "is the registry healthy" →
alpha_zoowithaction=health— surfacesloaded,failed, and per-error reasons. - User uploads
my_factor.csv→factor_analysis(zoo tools are for prebuilt alphas only).
Zoo Inventory
| Zoo | Description | Approx. count |
kakushadze101 | Formulaic alphas from Kakushadze's 2015 paper. Mix of momentum, reversal, volume, and microstructure. | ~101 |
gtja191 | Guotai Junan 191 alphas — A-share focused cross-sectional factors. | ~191 |
qlib158 | Microsoft Qlib's 158 alpha factors — features tuned for ML pipelines. | ~158 |
classical | Fama-French 3/5-factor + Carhart momentum. | <10 |
Counts are nominal; check alpha_zoo action=health for the live count currently loaded.
Constraints
- No per-stock per-date factor values are surfaced to the agent. IC results are aggregate stats (mean / std / IR / positive-ratio); the HTML report shows top-N by IR plus formulas, never the underlying panel.
- Lookahead is banned in the operator set.
delta(df, d)requiresd >= 1; the negative-shiftRef(df, -n)form does not exist. Seedocs/alpha-zoo/spec.mdfor the full operator catalogue. - Universe loaders may not be wired for every market yet. When
alpha_benchreturnsuniverse loader for X not yet implemented, that's the W2 scaffold — the universe is recognised but the data pull lands in W4. - Do not expose absolute filesystem paths in agent output. The bench tool writes to
~/.vibe-trading/reports/by default; refer to it by that shorthand, not by the resolved absolute path. alpha_zoois read-only.alpha_benchwrites a single HTML file per run — no scratch state elsewhere.
Common Pitfalls
- Filter mismatch on
list_alphas: theme / universe must match the alpha's declared metadata exactly (e.g.equity_cn, notcnorchina). - Calling
alpha_benchwith bothalpha_idandzooset — they are mutually exclusive; pick one. - Empty registry (
loaded=0) means no zoo modules are populated yet; treat it as "zoos pending W3 porting" rather than a bug.
Reference
- Operator catalogue:
docs/alpha-zoo/spec.md - Registry contract:
src/factors/registry.py(frozen; do not modify) - IC / layered NAV math:
src/factors/factor_analysis_core.py
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

