QuantSkills: Community Directory of Open-Source Quant Skills and AI Agents

What QuantSkills Is and What It Is Not

QuantSkills is a navigator repository for the QuantSkills GitHub organization, maintained at github.com/quantskills/quantskills. The README describes it as a discoverable, installable, verifiable, and shareable community catalog of quantitative skills and agents.

The project was initiated by PandaAI, but the README states clearly that it is not PandaAI's official certification and does not constitute investment advice. Contributors provide inspectable capability packages with stated boundaries; users are expected to verify them according to the documented data sources, assumptions, and risk limits.

The catalog itself is a static snapshot of metadata. The README states that the snapshot only lists public metadata, does not represent quality endorsement, revenue promises, or a guarantee of production readiness. This is a research and discovery directory, not a marketplace with validated products.

How the Catalog Is Organized

The catalog uses 10 categories covering the full quantitative workflow. Category 01 is data interfaces and warehouses, covering data sources, connectors, market data governance, and point-in-time data quality. Category 02 is the factor research toolkit, the largest category with 44 assets, spanning factor idea generation, factor synthesis, factor screening, and factor evaluation. Category 03 covers market and instrument analysis across A-shares, Hong Kong and US stocks, ETFs, futures, options, and macro cross-asset signals.

Categories 04 through 07 cover risk monitoring and alerting, strategy backtesting and trading tools, research models and paper replications, and research validation tools including lookahead bias detection, walk-forward testing, and signal stability checks. Category 08 covers news search and knowledge analysis. Category 09 covers AI agents for research, monitoring, risk, and execution. Category 10 covers infrastructure templates.

As of the 2026-09-19 snapshot, the catalog contains 214 assets across these 10 categories. Only 1 of those 214 assets has a published public endpoint. The remainder are listed as pending maintainer review with no public endpoint.

How to Find and Use Skills from the Registry

The main entry points are the navigator repository itself for browsing, and the registry repository at github.com/quantskills/registry for submitting metadata. The interactive catalog is at www.quantskills.ai.

Each asset in the catalog links to its own GitHub repository within the quantskills organization. For example, skill-pandadata-warehouse is a separate repository with its own README, inputs, outputs, and interface status. The catalog entry shows a summary, the primary workflow stage, what inputs the skill expects, and what outputs it produces.

To contribute a skill, the README points to skill-template at github.com/quantskills/skill-template. To contribute an agent, it points to agent-template at github.com/quantskills/agent-template. Contributors are required to document data sources, assumptions, parameters, limitations, and risk boundaries before submitting to the registry.

Workflow Map and Pipeline Stages

The README defines a 14-stage workflow map that skills and agents are organized around. The four high-level groups are data foundation (data-ingestion, data-quality), research signal (feature-engineering, factor-generation, factor-screening, modeling), portfolio validation (portfolio-construction, backtesting, evaluation, risk), and monitoring and trading (monitoring, execution, reporting), plus an orchestration stage.

Each skill in the catalog is tagged with its primary workflow stage. The skill-pandadata-warehouse asset has data-ingestion as its primary stage and produces market-bar output. This tagging allows users to filter the catalog by pipeline stage when looking for components to fill a specific gap in their workflow.

The workflow map positions QuantSkills as a composable system. Rather than finding a single end-to-end tool, users are expected to assemble a pipeline from components at different stages. This makes the catalog more modular but also means finding a compatible set of skills for a full pipeline requires more research than using a self-contained framework.

Interface Status and the Pending Review Problem

Every asset in the catalog shows an interface status. As of the 2026-09-19 snapshot, 213 out of 214 assets are labeled as pending maintainer review with no public endpoint. Only one asset, skill-pandadata-warehouse, has a published endpoint.

This is the central limitation of the catalog in its current state. A skill with no public endpoint may still work when cloned and run locally, but users have no way to verify this from the catalog entry alone. Some skills may be early-stage contributions that have not yet been validated; others may simply need the maintainer to complete the review process.

For a user looking to integrate a skill quickly, this status means the actual effort is similar to evaluating any unfamiliar open-source repository: clone it, read the code, test it against your own data, and verify the documented assumptions hold. The catalog provides discoverability; the validation work remains on the user.

Limitations and When to Use a Different Tool

QuantSkills is not a backtesting engine, a data vendor, or an execution system. If you need a standalone backtesting framework, tools like Backtrader or zipline are self-contained Python libraries where you write a strategy and run it against historical data without needing to discover and assemble components from a catalog.

The catalog is specific to the kinds of tasks AI agents perform in quantitative research. If your workflow does not involve AI agents or you prefer a traditional programmatic approach to quant research, the QuantSkills catalog adds an extra layer of indirection without clear benefit.

The repository has no stated license (shown as unknown in the repository metadata). Individual skills in the catalog have their own licenses in their respective repositories, but the navigator repository itself has no license declaration. This creates legal uncertainty for anyone who wants to incorporate the catalog content into a commercial product or redistribute it.

Maintenance and Community Structure

The QuantSkills organization has several entry repositories: the navigator at github.com/quantskills/quantskills, the registry at github.com/quantskills/registry, templates for skills and agents, a join repository for community discussions, and community rules at github.com/quantskills/join. The last push to the navigator was on 2026-09-27, suggesting active maintenance.

The community guidelines are documented in the join repository. Discussions happen through GitHub issues and pull requests in the registry. The README encourages contributors to explain data sources, assumptions, parameters, limitations, and risk boundaries for each submission.

The JavaScript-based build tooling in the navigator repository, shown in package.json, runs a build script (scripts/build.mjs) and tests (tests/*.test.mjs). The catalog page at www.quantskills.ai is generated from the metadata stored in the repository.

Editorial conclusion

QuantSkills is a useful starting point for quantitative researchers who work within an AI agent workflow and want to find reusable, inspectable components for specific pipeline stages. It is not a trading system, a backtesting engine, or a production data feed. Before using any listed skill or agent, verify the interface status shown in the catalog: most assets are listed as pending maintainer review with no public endpoint, meaning they require local setup and independent validation. The repository has no stated license, which is a real constraint for commercial use.

Frequently asked questions

Official sources

1. Issues
2. quantskills/quantskills on GitHub
3. README